Voice Communication in Business Volume 2
Essays on telecommunications, 1981-2002

This one never appeared in print, but I wrote and rewrote it, along with They’ve Got Your Number, about eight times during '92 and '93, hoping it might be of use to others. What I wanted to show here was just how complicated numbering plans had already become in a business context.

Looking back from 2002, when the Internuts are busy trying to make IP-PBXs and Voice over IP phones, I can't help but wonder how they will avoid the mistakes of the 1975 PBX designers who thought the customer would accept anything because it used new technology, even though it couldn't provide most of the traditional services and features users had come to expect.

Others Have Your Number, Too
(1993)

The North American numbering plan is in trouble: we are running out of numbers right on schedule, and no matter which of several solutions is chosen, there will be considerable confusion and resistance. If you like that one, however, you'll love the much less appreciated numbering plan problems inherent in PBXs. PBX numbering plans are infinitely more difficult than those for the public network, and have as many approaches as there are vendors.

National Vs. PBX Numbering Plans

To select a specific item, that item must have an identity or "address" that is unique. Otherwise, the selection process has no way of knowing which of two or more otherwise identical items is to be reached. One might suppose that the ten billion telephone numbers implied by a ten-digit numbering plan would provide enough addresses to insure the uniqueness of every phone and every other terminal; that this is not the case is all too evident.

It has been traditional for the telephone numbering plan to help locate the physical disposition of telephones to simplify finding them, wherever they may be. Thus we have the country divided up into areas, each with a unique three-digit area code, and within an area code, we have three-digit office codes to identify the specific switch which serves the telephone we want. Each office code, in turn, supports 10,000 station numbers, based on four decimal digits; an office code and a station number make up a directory number. Note that modern central offices switches can serve over 100,000 lines; thus an office code defines a switch, but a switch is not very helpful in defining its ten or more non-sequential office codes.

The general idea, when Direct Distance Dialing (DDD) took over in the 1950s, was to have customers always dial a seven digit directory number to call in the same area code, and dial a ten digit number to get any other phone in the United States and Canada. From the customer's point of view, the worst thing about DDD was the need to dial seven digits for a local call. In most smaller cities, using Strowger Step by Step (SXS) switches, it had been common practice to dial only four digits, perhaps adding a fifth digit as the number of phones grew.

When the fifth digit was added, telephone users quickly discovered that, very often, they could still reach other people whose telephone number had the same new first digit as theirs by dialing only four digits as before. But with the coming of DDD, this trick no longer worked. They HAD to dial seven digits. And, to add to the difficulty, exchange names, using letters on telephone dials, were abandoned by the telephone company in favor of ANC, or all number calling. The furor over ANC and all the numbers required by DDD has even today not died down completely. A certain element of revenge has been extracted, however, by companies using dial letters to make their 800 numbers easy to remember: 800 Business Communications Review 1234 and 800 LIBRARY are examples known to us all.

The problem, of course, was that common control switching systems had to be used to implement the inter-switch routing required by DDD on a nationwide basis, and common control switching HAS to have a "uniform numbering plan." With SXS, a uniform numbering plan was not needed; one could have as many or as few digits as were required to reach a destination, and when the last digit was dialed, the calling party was actually connected to the called party. With common control, the user did not dial digits directly into the switches as with SXS; rather, digits were dialed into a register. When the customer finished dialing, the register would flag the common control, and common control would use the contents of the register to look up the proper routing. Then the system would START to set up the connection.

The need for a uniform numbering plan came from the register's need to know when to flag the common control. With a uniform numbering plan, if the caller always dials the same number of digits, the problem ceases to exist. Unfortunately, there has never been a uniform numbering plan in the real world. With DDD, the caller had to dial seven or ten digits, but there were always other problems such as dialing 0 for operator and 911 for emergency. To deal with such exigencies, three additional techniques are available: time out and if no more digits come in, go with what you have; require the user to send an "end of dialing digit"; and use a technique called "pre-translation."

With pre-translation, the register (and/or common control) examines each digit as dialed to estimate the number of digits still to come, and to see if enough digits are available to take action. To make pre-translation work, the national numbering plan was designed so that both office codes and area codes did not begin with a 1 or a 0, that office codes used only 2-9 as their second digits, and that area codes used only 0 or 1 in the same spot. If pre-translation showed the second digit to be a 1 or 0, the common control knew eight more digits would (probably) follow, while if the second digit was 2 through 9, there would be only five more.

These limitations have created today's numbering plan problems. It should also be noted that small communities seldom fill up the office code assigned to their switch, while large switches have unused station numbers within their various office codes to allow for growth and churn. Between limiting office and area codes to simplify pre-translation, and not always being able to assign all possible station numbers within an office code, literally billions of potential telephone numbers do not exist, and we have a shortage.

The solution being implemented is to allow both area codes and office codes to have any of the ten digits in the middle, retaining seven digit dialing within an area, but requiring the caller to dial 1 followed by an area code to escape to the outside world. Checking for an initial 1 would make pre-translation easy if it weren't necessary to deal with 0 for local operator, 00 for long distance operator, 411 for directory assistance and 911 for emergency, and a variety of codes involving 1 and 0 to identify a particular long distance carrier, an overseas call, etc. Thus time-out must be added to pre-translation.

The telephone network seldom requires the public to use an end of dialing digit, but prior to the coming of Common Channel Signaling, KP and ST were used to mark the beginning and end of a number transmitted by operators or senders. This limited outpulsing to the digits the distant office actually needed (you didn't spill forward an area code if the trunk being used came out in that area, for instance), minimizing call set-up time.

In passing, it should be noted that one of the principal advantages of common control systems is their ability to do "translation" as well as pre-translation. With translation, each matrix port is defined by an "equipment number" (cabinet, shelf, slot and circuit in slot, for instance) for the convenience of the machine, while users continue to dial conventional directory numbers; the common control has a look-up table that allows any directory number to be assigned to any equipment number, giving the system enormous administrative flexibility.

With a lot more area codes, and a few more office codes within each area code, the DDD numbering plan can probably continue for some years. However, the numbering plan in the public network is relatively simple compared with those needed by PBXs. The great majority of telephone numbers in the national numbering plan address residential and small business telephones. As far as the numbering plan is concerned, each number identifies a particular port on a switching matrix in some local Central Office switch, or perhaps a group of lines to access a PBX or ACD.

Contrast this with a PBX numbering plan. First, a PBX today, being a common control system, must have a uniform numbering plan. However, for internal calling, the equivalent of an office code is not used. As a result, two, three, four and sometimes five-digit systems are common. But internal calling is only a small part of PBX activity. Calls have to be placed to the outside world, and often, directly to other PBXs. Thus access (escape) codes such as dial 0 for the console attendant, 9 for outside (to reach a group of CO trunks), or 8 for the company network (tie-trunks) are used with pre-translation.

In the hospitality industry (hotels and motels), single digit dialing is continued for frequently called functions such as front desk, cashier, room service, etc., often requiring an escape code to reach room phones. "Just dial 7 and the room number" is simple enough for pre-translation until it is called upon to separate room 235 on the second floor from room 2354 on the 23rd floor. PBXs used in apartment buildings sometimes use an escape code for INTERNAL calls to other apartments (often * or #) so that tenants can dial most numbers into the PBX's register as though they were dialing directly into the CO.

Business customers often have a variety of special connections to local and long distance carriers which may require multi-digit access codes. In addition, modern PBXs have a number of features accessed by dialed codes. To make all this work, pre-translation, time-out and an end-of dialing signal may all have to be used, even when care has been taken not to allow extension numbers to have the same first digit(s) as feature codes or access codes.

But such limitations on the numbering plan are relatively simple. The problem gets sticky is in dealing with actual business telephones themselves. In general, business telephones come in groups, emphasizing the way people work together. Thus single-line instruments, typical of those found in residences, are not suitable; rather, multi-button phones are often required to allow one phone to have direct access to several extension numbers, one extension number to appear on several phones, and a variety of features, meaningless on a single residential line, to be interfaced effectively. Establishing the relation between dialed number, port(s) on a switching matrix, and telephone sets is far more complex than it looks, and each manufacturer has different ways to do it.

Business Phones: The 1A2 Heritage

Serving business needs was standardized years ago with multi-button phones in key systems, usually referred to as 1A2. With a key telephone set, one phone could pick up several different lines, one line could be picked up on several different telephones, and the human interface, even though limited to lamps and buttons, developed over 30 years into something so effective that even executives could use it without formal training. A typical procedure was to combine PBX line-hunting with the appearances of the several lines on 1A2 key telephones so that a call finding a line busy could go to the next and still be answered at the same group of telephones by simply pushing down a different button.

Key systems had a control circuit called a Key Telephone Unit (KTU) for each extension number they served; KTUs were mounted in a Key Service Unit (KSU); each extension number had a single pair of wires from its KTU to a port on the switching matrix, but each phone was connected to the KSU by a fat cable that provided several pairs for each button. The familiar six-button key telephone set, for instance, was connected to the KSU by a 25 pair cable. To associate a given button with a particular extension, cross connections were made in the KSU between the fat cable and the extension's KTU. Note that the entire operation of the key system was invisible to the PBX (or Centrex). As far as the switch was concerned, it was connected to nothing but single line telephones, often in a hunt group.          Although this technology was pretty well standardized by 1950, the PBX numbering problem began to show when several extension numbers and matching matrix ports were associated with several telephones. There was no a priori reason to think of any particular key telephone set as being associated with any particular port (or line). What was usually done when a three-line hunt-group appeared on three phones was to arbitrarily give each phone the identity of one of the lines; at least for inventory purposes, back when the telephone company owned everything, this was perfectly satisfactory.

When only two lines in hunt appeared on three key telephone sets, there weren't enough extension numbers to provide set identification. What was done was use the extension numbers to identify two phones, and use either number followed by an "A" to identify the third phone. Thus we would have extensions 236 and 237 appearing on telephones 236, 236A, and 237. And if line 236 was busy (at any of the three phones), the PBX would hunt to line 237 (if it was free) and ring it. The call could then be answered by either of the two remaining phones, or the first one if it put its existing call "on hold."

This is an instance of an "A station," a phone without a line of its own. Needless to say, large key systems might have several A, B and C stations when there were many more sets than lines. The inverse situation was also common: frequently, there would be more lines than phones in a group, a situation typical of Automatic Call Distributors. Even in a conventional office, however, a busy executive might have four lines in hunt, available on the executive phone as well as on the associated secretarial phone. The secretary would answer all calls, put them on hold, and announce them to the boss via an intercom channel. Both boss and secretary could hop from one line to another by using the hold button on the current call and a line-select buttons to connect to a different call. The PBX didn't care if each line had a phone of its own, as long as it had a KTU.

It often happened that certain "private" lines might appear on the executive phone but not on the secretarial phone. This posed a new problem when the executive did not want to be disturbed even though the secretary had no lamp and button for the line actually in use. A common solution was to provide an extra contact on the hook-switch in each phone which could light a lamp on another phone. This indicated that a specific phone was in use, but not necessarily which line it was using.

As can be seen, the purpose of a key telephone set was to allow access to several different lines. However, the most common use made of this capability was for a secretary or assistant to place outgoing calls and screen incoming calls for a principal. This required an intercom between the two, so that the boss could tell the secretary who to call, and the secretary could tell the boss who was on which incoming line. This intercom also required each set to have an identity independent of the PBX or CO lines the set served.

In small companies, where a key system allowed a number of telephones to have access to a much smaller number of CO lines, it was not unusual for the receptionist to answer incoming calls with the company name, obtain the name of the called party, put the calling party on hold, and use an intercom to tell the called party who was calling and which line they were on. Such large intercoms again required the telephone set to have an independent identity.

As these examples demonstrate, the relations among matrix ports and telephone sets are ambiguous, and telephone sets may have to have an identity of their own, independent of the extension numbers (or directory numbers) they serve. That is, ports, numbers and telephone sets are quite different entities. However, in electromechanical SXS systems (the great majority of PBXs prior to 1975), the switch neither knew nor cared about these problems. The key telephone system was completely invisible.

Enter Stored Program Control

The coming of stored program control led, of necessity, to new possibilities. Because of the high cost of computers in the 1960s, stored program control was originally used in COs where the cost could be pro-rated over a very large number of lines. Even so, electronics at that time was vastly more expensive than traditional switches, so new features had to be developed to justify this higher cost. Hunting, for instance, was improved: no longer did hunt groups have to consist of sequential directory (or equipment) numbers, hunting could take place after N rings as well as on busy, many numbers could hunt to a single number if desired, and "circular" hunting was added to the "terminal" hunting which the motion of SXS switches did naturally. A variation of hunting called call-forwarding allowed the user to establish hunting patterns on a "follow me" basis; one could now go next door for an evening of bridge with the neighbors and still get one's calls.

New features, intended for the residential market, included call-waiting and consultation/conference (which later, in PBX and Centrex systems, incorporated the ability to transfer calls; just add a third party to a connection and drop off). Both of these features required the use of a switch-hook "flash" so that residential single-line phones already in place could use them. With call-waiting, the flash would put an exiting call on hold and connect a new call, while subsequent flashes would permit alternating between the two connections. For consultation/conference, a flash put an existing call on hold and gave the caller dial tone. A new connection could then be dialed for private consultation, with a second flash producing a three-way conference These features allowed one line to be used for two calls, call-waiting handling a second incoming call and consultation/conference allowing a new outgoing call while keeping an earlier call on hold. (It should be noted that dropping a consultation connection and returning alone to the original call or, on a transfer, being sure the transferring party had dropped off were major privacy problems in many systems).

The presence of large quantities of (relatively) inexpensive memory also led to repertory dialing, variously known as abbreviated dialing, speed calling, etc. With repertory dialing, the customer could now substitute one or two digits to complete a local or long distance call. Rep dialing interacted with the national numbering plan, however, by starting with the same digits used by area or office codes. When two-digit codes were used to identify the caller's thirty most called numbers, the switch had to be able to differentiate between a 32, meaning George on 555-4726, and 32 as the first two digits of some office code. A time-out was used, making the term Speed Calling somewhat suspect. However, a user in a hurry could append an end-of-dialing code (# on a TouchTone phone) rather than wait.

By 1975, computer prices had dropped enough to allow them to be used in PBXs; even so, their cost was such that even more new features had to be added to consultation/transfer, repertory dialing, advanced hunting, and call waiting to create the impression of a bargain. Call pickup was one of the most interesting. It came in two forms, "directed" and "group." With group call pickup, a person could answer a ringing phone in his or her pickup group by dialing the group pickup code. Directed pickup required dialing a different pickup code followed by the number of the ringing phone. Each variation had different applications.

These features, and several hundred others, all based on adding value to single-line phones, misled system designers, few of whom had ever heard of 1A2, into believing that residential telephones could meet all user needs in a business context. But call-waiting was mutually exclusive with hunting (or call forwarding on busy), making it impossible for a secretary to answer the new call, while consultation/conference/transfer required more agility than most people possessed and had built-in privacy problems as well. To add to the fun, conference/transfer worked differently on different switches, and often in different ways within a given switch, depending on whether the user had made or answered the call in progress, or the call was inter- or intra-PBX. But the most important failure of infinite features on single line sets was their total inability to provide secretarial assistance with executive calling.

When customers refused to give up 1A2 behind their new PBX or Centrex systems, tricky problems developed. For instance, a call-waiting tone applied by the PBX to a line put on hold by the 1A2 would not serve its function. But even more interesting was the interaction of two ways to put a call on hold: if a secretary flashed a switch hook to activate a consultation/transfer or call-waiting feature, and then operated the 1A2 hold button to go to the 1A2's intercom channel or another line, it was possible to have an "invisibly held" call which did not show up on the 1A2 lamping. AT&T was particularly concerned with this problem, which could leave calls in limbo.

There were a few designers, particularly at Northern Telecom, who understood the needs of business communication. As an intrinsic part of SL-1 development, they designed electronic key telephone sets (EKTS) to provide the features of 1A2 while taking advantage of much more effective hardware. "Skinny wire" EKTS eventually caught on with many manufacturers, and today fill a very important need, particularly where boss- secretary operation is required. But here the numbering plan on stored program PBXs reached a new level of complexity. Skinny-wire telephones were based on a single talk path and an independent signaling channel between set and line card. The general idea was for the set to send information to the PBX via the signaling channel to tell the PBX to use its switching matrix to make connections formerly made in a 1A2 set with a line select button. And also via the control channel, the PBX told the phone to operate various audible and visible displays to emulate 1A2 for the benefit of the customer.

For instance, the user could answer any of several calls by pushing down the appropriate line-select button as with 1A2. However, the feature actually at work was "directed call pick-up" with a better interface for the user. Directed call pick-up on single line phones could be made to work if (a) the customer knew a phone was ringing unanswered, (b) knew the extension number of the ringing phone, and (c), remembered the feature code to dial before dialing the ringing extension number. It was much easier to simply look for the line-lamp being flashed and depress the associated line-select button. Either way, however, the switch used its matrix to connect the call to the talk path of the answering phone. Similarly, pushing "hold" and then a different line-select button to get dial tone was much easier (and more reliable) than trying to activate consultation hold with a switch-hook flash.

Almost all EKTS have a one-to-one relation with a matrix port to simplify synchronization and signaling between set and switch; few systems (the Northern Telecom DMS CO switches are an exception) can bridge two or more EKTS across the same line as is common with 2500 sets. Indeed, "bridging" has been reinvented as a new feature when the same extension number appears on buttons on two or more sets. One set can join a conversation already established by another using a particular extension number by simply pushing the line pick-up button for that number, but conference circuitry in the switch does the combining.  The one-to-one relation between an EKTS and a matrix port is perfectly logical, but the relation of both to an extension number is more complex. As with 1A2, there is no compelling reason to associate a set with any particular one of the extension numbers on its buttons. However, the switch can no longer ignore the multi-button set and pretend each of its ports goes only to a single line phone as with 1A2. The switch controls the operation of the EKTS it serves, and if extension 345 appears on four different sets, the switch has to deal with those four sets on four different matrix ports for everything that happens on a call to "extension" 345. Similarly, each set has to be controlled for calls that come to all of the extension numbers on its buttons.

One way to deal with the problem was to insist that each EKTS have its own "prime line," an extension number that "belonged" to it. Thus every port would be guaranteed a directory number; but if every set had to have a prime line number, typical 1A2 installations where three "lines" served six telephones could not be implemented. Then, too, the inverse problem had to be considered; could you have extension numbers without sets? If so, some aspects of program organization could be simplified, but if not, the need for pilot numbers for hunt groups, parking orbits, and certain functions unrelated to hardware sometimes led to closets full of connected but unused telephone sets, served by unnecessary line cards, to allow such numbers to exist.

AT&T's System 75 and Definity, Generics 1 and 3, still require each telephone set to have a prime line, but Generic 3 has made it possible to have prime lines without sets of their own. Why it took so long to achieve this is unclear, but the solution arrived obliquely, to say the least. G3 has a new feature called "administration without hardware." It's purpose is to allow all the software administration for a new addition to be completed before the hardware is installed so that the hardware can be checked out without waiting for all the related programming to be done. However, when it became evident that the new hardware did not have to be installed at all, pilot numbers became possible, you could get an extra extension number on your EKTS, etc. Thus Definity G3 will let you have six numbers on three phones, but you still can't have six phones with only three numbers.

Both Rolm and Northern Telecom have simply expanded translation in a straightforward way so that numbers dialed by users can be identified with one or more ports, or no ports at all, as required. In particular, the system memory related to a dialed number can cause the switch to do what needs to be done, even if no hardware at all is associated with it.

Variations On A Theme

System designers have found the temptation to "improve" on 1A2 almost irresistible. For instance, 1A2 emulation would require a line pick-up button to produce bridging to a call on an extension number already busy, and to signal directed call pickup for a ringing extension. But if the extension number is idle, there are two possibilities. For true 1A2 emulation, pushing the button should give you dial tone so that you can make a call ON that extension number. However, another possibility would have the button activate a repertory dial function to let you place a call TO that extension number.

Call billing, it might be argued, should be based on the telephone (and user) making the call rather than on an extension number that is a prime line on some other phone. However, when a secretary sets up a call for a boss, billing to the boss's line is exactly what we want, particularly when the secretary happens to serve several different principals. Thus something can be said for true 1A2 emulation. However, pushing a button to call somebody else in one's work group has a lot to recommend it. Various systems such as the NEAX 2400 have two sets of buttons, one just for repertory dialing, and the other for line pickup.

Another opportunity for improvement concerns ringing. Because each EKTS has its own matrix port and is independent of all others, the alerting function can take on a variety of forms: extension numbers can have their lamps blink to give a visual indication of ringing at all sets, but on some sets, the audible ringing signal sounds immediately while on others it may be delayed or deleted. For example, the secretary's phone might ring at once, but the boss's phone only after three rings; this keeps the executive office quiet but provides a safeguard in case the secretary had go to the Xerox machine for a moment. Alternately, a staff member's phone might ring immediately, blink but not ring at the phone of the departmental secretary, and only start to ring at the message center after a delay. There is obviously an advantage in having the same number rung in different ways on different phones and, of course, different numbers on the same phone rung in different ways. 1A2 ringing was more limited, but that is no reason to forego such convenience now that it is possible.

Although 1A2 emulation is simple, logical, and takes advantage of what the customer already knows, some systems do not even attempt it. And even when a system can emulate 1A2, the limited number of buttons on each set suggests that there will be many extensions for which pickup, transfer, and conference will need a different approach. For example, call pickup for a large group can have a single button and lamp on an EKTS; when any line is ringing in the group, the lamp blinks. Pushing the button performs the same function as sending the group pickup signal. When LED displays first became available, the ringing number could be shown; Rolm's ETS sets did this, and in a later version using an LCD display, added the name associated with the extension number. These displays were also used with call forwarding; a message center phone might show that a call is being forwarded from Ext. 235, intended for Miss Smith, because her line is busy. Rolm's current electronic sets, the ROLMphones, preserve these capabilities but increase the number and flexibility of line/feature buttons required.

It is easy enough to have an EKTS with nothing but a prime line number, a bunch of feature-activation buttons, and an alpha-numeric display. By taking advantage of the display, it is possible to identify and answer specific ringing lines, put a number of calls on hold and retrieve them selectively, and do many of the other things that rival 1A2 operation. However, user training, generally unnecessary with 1A2, may be required.

Hardware With Numbering Plans

Stored program control freed the business communication manager from the limitations of SXS when numbers no longer had to be consecutive to hunt, but even more important, there was no longer a need to "fill up the 200 level" before buying new switches to add numbers in the 300 level. A 60 line PBX could have four-digit numbers if desired, using the first two digits to identify department and work group within that department. That is, the PBX numbering plan could be used as a tool by the communication manager; in this example, CDR could sort on the first digit to get department telecom costs, and on the second digit after each first digit to get work group costs.

Although communication managers immediately saw the value of such flexibility, they were not always able to exploit it. Two features, available long before stored program PBXs came on the market, continued to limit stored program's potential flexibility: Direct Station Selection (DSS) and Direct Inward Dialing (DID). DSS added a large panel to the attendant console on which was located a button and lamp for every extension in the PBX. The purpose of DSS was to speed up the completion of incoming calls by allowing the attendant to push a single button to tell the system to connect an incoming trunk to the station selected. DSS panels for up to 200 lines were designed; obviously, the lamps and buttons had to be numbered consecutively and in an orderly manner so that a called extension number could be located easily. This limited numbering plan flexibility, and of course there could be no numbers not on DSS.

DSS was introduced in the late '60s and early '70s, before DTMF detectors became cheap enough to put in PBXs. With a rotary dial, it took the console attendant forever to dial the called extension into the common control, causing other incoming calls to be backed up in queue. Compared with a dial, DSS worked like lightning. But DSS in general did not hunt. To put a secretarial line in some logical place so that the console attendant could easily find it when the executive line was busy reimposed something very like sequential numbering for hunt groups and jacks on a manual switchboard. Note that DSS buttons did NOT act like line select buttons in 1A2; rather, they acted like buttons on a repertory dialer. Before the coming of inexpensive electronics for signaling, individual wires ran between each lamp/button and the switch, and 300 pair cables to the console were common.

Until about 1975, large PBXs used cord boards rather than consoles to support their electromechanical switches. When electronic PBXs for a thousand or more lines became available, cord boards were unthinkable and consoles had to be provided. DSS for large systems usually had only 100 lamps and buttons, and used additional buttons to assign the array to a particular hundreds group. When the attendant had to push one button to identify a group of 100 lines, and then push a second button to specify a line in that group, it was quicker to push three or four buttons in succession on a standard 4x3 signaling pad.

In console service, DTMF was not recommended because an attendant could, after a little practice, dial in numbers faster than the talk-off protection of a DTMF receiver could accept them. Rather, an early form of digital signaling, unsuited for use with user telephones, was often employed. Between DSS and digital signaling from keypads, consoles became firmly entrenched as a necessary part of common control PBXs. Even when the power of EKTS was equal to that of a console, many PBXs still required special consoles with special wiring to the switch and special training for console attendants. There was no longer any reason to make consoles different from EKTS, or to force PBX extension numbers to conform to the DSS panel layout, but old habits die hard.

Stored program control had some new tricks to offer DSS. Not every station user needed DSS; why not make it available only where it made sense? With lamp-button combinations which could be assigned as desired to specific extensions, alphabetical DSS for people receiving large numbers of calls became practical. Solid State Systems was one of the first to do this.

Turning now to direct inward dialing, DID continued to force customers to use numbers in compact blocks because the telephone company only translated on enough digits to pick the block, not the individual number. The scarcity of numbers in the national numbering plan makes hoarding DID and Centrex numbers anti-social, and telcos do not like to reserve additional blocks of numbers for a client until the first block is used up. This sometimes leads to non-contiguous numbering blocks for a given client, or massive number changes for all extensions when growth forces a customer into another office code.

A major advantage of PBX DID over Centrex is that in Centrex, all extension numbers are part of the national numbering plan, while with PBX DID one only need use "outside" numbers where they serve a purpose; PBX extension numbers can be used for other lines. In a PBX, DID numbers can hunt or forward to non-DID numbers, allowing second and third lines for a busy person to be internal extensions, large hunt groups or ACD groups needs only one DID pilot number for outsiders to call, and people receiving only a few outside calls can be accessed via the company's directory number, transferred by the console attendant.

With Centrex or DID, the final digits of the outside number are the equivalent of an extension number. Thus when using inside extensions as well as outside Centrex numbers, care must be taken not to have the two groups overlap. For instance, if the DID numbers are 555-2xxx and 555-3xxx, extension numbers can be 4xxx, 5xxx, etc. When feature codes are used, they, too, must not overlap DID numbers, extension numbers, or access codes. Thus if one dials 9 to get outside, 8 to reach the company tie-trunk network, and all feature codes start with 7, DID and extension numbers in the 7xxx, 8xxx and 9xxx blocks are excluded. If the only block of DID numbers available from the telco is in the 7xxx series used by features, vast user retraining will be necessary. A number of systems use * and # to identify features, which works fine with DTMF phones but not so well with rotary dial.

DID is not the only way for an incoming trunk to reach a particular line; Direct Inward System Access (DISA) and Direct Department Calling (DDC) are also frequently used. With DISA, the PBX responds to ringing on an ordinary CO trunk by answering and giving the caller PBX dial tone. The outside caller can then use DTMF to reach any internal extension, dial-accessible trunk group or feature. Today, many PBXs use recorded announcements of varying degrees of complexity rather than PBX dial tone, and provide callers with voice menus for selecting inside users or functions. This improved "dial tone" is called an "Automated Attendant."

DDC allows outside calls arriving on specific CO trunk groups to reach certain internal extensions or hunt groups, bypassing the console attendant and delivering the call to the first free member of the group. With both DISA and DDC, the PBX extensions are identified by their own extension numbers and not the outside number the caller dials to reach them. Typically, trunks reached by dialing a regular telephone number are arranged to go to an internal PBX pilot number, a space in memory which translates to a hunt group composed of several internal extensions. Only with a version of DDC where a single line reaches a single extension number, called a "Direct In Line," will the phone or button be labeled with the outside number. DILs are generally used to allow private outside lines to appear on proprietary electronic sets which have signaling incompatible with the CO; they provide PBX features such as conferencing, call forwarding, etc., on the private line. Note that 1A2 key telephone systems could terminate both PBX and CO lines directly (both the CO and PBX used similar dialing and ringing), allowing the earlier equivalent of a DIL to be independent of the PBX but denying it the availability of PBX features.

Numbering For Electronic Key Systems (PBXs)
Behind Centrex

What happens to the numbering plan when we have a PBX behind Centrex? Today, this occurs because PBXs may be needed to provide business features which Centrex does not offer. Such PBXs are often called electronic key systems or "hybrids," but they are still PBXs; CO trunks (lines) from the Centrex switch terminate on trunk circuits at the PBX, extension phones terminate on PBX line circuits, interconnection is made by a switching matrix, and a computerized control makes the whole thing work. Thus, as Dick Kuehn has pointed out, a customer either has a large PBX supporting electronic phones directly or Centrex with a large number of small PBXs supporting electronic phones. (As mentioned above, however, the Northern Telecom DMS-10/100 supports a series of proprietary analog electronic phones controlled by a separate digital signaling channel, which a customer can buy separately where key telephone features are needed. Someday, ISDN will make standard digital telephones available, and Centrex will be able to do what PBXs have done for almost two decades).

The terms hybrid and electronic key systems were invented by lawyers and economists to allow PBXs to be sold in states where the cost of PBX trunks was vastly higher than the same facilities used as customer lines. Although reality is seldom of interest to a lawyer or economist, a definition was needed to permit these trunk savings to be realized. One attempt held that with a key system, one could choose a line (trunk) to the serving switch by pushing a button, while with PBXs, one could not. Although silly, this does suggest an aspect of truth in connection with numbering plans. A PBX or a CO switch defines the extension numbers it serves, while a key system does not. A key system need not concern the user with the identity of its "line-side" ports; its phones have pick-up buttons identified by line numbers on the serving switch. Clearly, an electronic PBX acting as a key system behind Centrex is simply a PBX that specializes in DILs.

Although 1A2 itself did not provide station numbering, intercoms, when used, sometimes did. Intercoms were typically quite limited, run only between a secretary and boss to announce screened calls. Each person had a signaling button and a buzzer; when one pushed the button, the other would hear the buzz. Various codes were often worked out: one buzz to pick up the phone, two to come into the inner office, etc. Unfortunately, few electronic key systems choose to offer this popular feature, finding "dial intercom" more profitable.

"Dial intercom" on 1A2 was a signaling trick; the caller would push the button to connect to the intercom channel and then dial the intercom station number of the called party. A small switch in the KSU would then operate a buzzer in the selected phone to tell the called party to connect to the intercom. For our purposes, the intercom station number corresponds to the extension number on a PBX. With an electronic PBX acting as a key system, where the signaling and talk channels to the electronic phone are used for intercom as well as other connections, intercom station numbers really are extension numbers.

Large 1A2 key systems, supporting many telephones, developed a variation of DSS. The key-set used as a console had an additional lamp/button panel, with one lamp and matching button available for every PHONE in the system. The lamp would show the busy/idle status of the telephone set itself; the button would operate the buzzer in that phone to tell the user to meet the attendant on the intercom to find out which line to pick up to get an incoming call (later versions used voice announce). Such key-system DSS buttons were usually labeled with the name of the station user rather than some "extension" number related to the intercom, but either way, each PHONE had a unique identity or address. Electronic key systems, taking advantage of their advanced signaling technology, can emulate this approach very well, and provide more satisfactory switched connections for intra-key-system intercom calling. Stations without DSS buttons can usually dial intercom (extension) numbers to make internal calls.

Because extension numbers on multi-button electronic sets are symbolic rather than real, and EKTS are independent of one another, variations in using telephone numbers are almost unlimited. Quite early, Northern Telecom provided an option to 1A2 on its SL-1 sets: mini-ACD. Say extension 237, part of a DID number 555-1237 appeared on six phones. Outsiders dialing that number would cause all six phones to ring, but as soon as somebody answered, the remaining five phones would be available for other calls. Only when all six phones were connected to different calls dialing 555-1237 would the system return busy. But clearly, this approach could not easily support the conventional boss-secretary pattern; once the secretary answered, the lamp on the principal's phone would go out, and the secretary would have to use transfer to complete the call.

EKTS had the ability to solve certain 1A2 problems. For instance, 1A2 line buttons had mechanical latches so the handset remained associated with the last button that had been depressed. An executive, hearing the phone's intercom buzzer, would pick up the phone, push the intercom button, and be greeted by the secretary who screened incoming calls. Upon being told which line a new call was on, the executive would push that line button and, more often than not, get dial tone.

What had happened, of course, was that the calling party had reached the executive's line and been put on hold by the screening secretary. When the executive picked up the phone upon hearing the intercom buzzer, it was still on the outside line from the last call. Going immediately to intercom disconnected the outside call, as the executive found out upon returning to the outside line. This scenario caused literally thousands of trouble calls, and ultimately led to 1A2 key systems adopting telephone sets that popped up any depressed button when the handset was returned to its cradle. (Such phones could not use their switch-hook to flash the connecting switch, and had to have a separate flash button for that operation).

Needless to say, it was too much to expect executives to modify their telephone habits to accommodate ivory tower system designers. However, EKTS on stored program PBXs were able, with a little effort, to accommodate the executives. A feature called "prime line preference" would cause an EKTS to reselect the set's prime line upon hang-up; when this prime line was the intercom, actually the PBX extension number on an electronic key system as opposed to a line number (on a PBX trunk port) defined by a Centrex switch, everything worked fine. On outside calls, executives would have to select something other than the intercom by pushing a button, but usually they would expect their secretaries to place or answer calls and then tell them which Centrex line the call was on. Note that such executives seldom allowed any line to ring in the sanctity of their offices; the only sound the phone would make was for the intercom, and that only when the secretary knew awakening the boss was permitted.

Another solution, of course, was to design the system so that a user could go from one line-select button to another without losing the call on the first button; it would simply go on hold automatically, as when the user flashed the switch-hook on single line sets. While widely used, this approach made it difficult for high-volume callers to hang up an old call by taking a new one. Obviously, you can't have everything.

Although call waiting and consultation/conference could, in print, allow one line to act as two, users had difficulty making these features work. Plessey of Canada, some years ago, put additional displays on the phone and used digital signaling to make such operations easier. This allowed one phone with one phone number to handle up to four simultaneous calls, with better signaling reducing user ambiguity.

AT&T's Horizon achieved this same effect by its efforts to avoid the "invisibly held" calls that had been discovered when 1A2 was used behind Dimension. Each phone had a prime line with THREE appearances, each labeled with the SAME extension number. This allowed, for instance, a new incoming call, which would have invoked call waiting on a 2500 set, to have a separate appearance where it could be answered just as though it were on a different line; with its own lamp and button, the call wasn't invisible any more. Similarly, such a button could be used to obtain a new dial tone for consultation purposes. AT&T has carried this multiple appearance concept through to Systems 75, 85 and Definity, and in these systems, electronic sets do not allow a switch-hook flash to invoke features.

To allow secretarial screening, call set up, etc., it became necessary for the principal's three prime line appearances to appear on the secretary's phone, where that phone had a prime line (with several appearances) of its own. A secretary serving several principals has to have a lot of buttons. However, this approach allowed the principal to have the equivalent of call-waiting which was not "invisible" to the secretary; both the boss and the secretary would be aware of the presence of a second or third incoming call, and could answer as the situation demanded.

Currently AT&T has backed off on the need for three appearances of a prime line, but if only one appearance is provided, the user of an electronic set cannot use call-waiting, conference or transfer the way someone with a 2500 set can. However, hunting and call-forwarding are available for incoming calls, and a different line appearance (perhaps the secretary's prime line) can be used for consultation.

Stored Program And Administration

The flexibility of stored program control was particularly effective when it was realized that it enabled system administration to be carried out from a relatively dumb terminal (such as the console or an EKTS). Some systems based the programming for a user on the matrix port (equipment number) associated with his or her phone, and used an administrative terminal to insert the appropriate prime line number, extension numbers and features for other buttons, etc. However, when the user moved to a different office, the whole process had to be repeated for the new matrix port, or a jumper change had to be made at the MDF (cross connect frame) so that wires from new office reached the old port on the matrix. In many instances, the old-fashioned jumper change proved to be an easier way to do translation than using a modern admin terminal. However, by making "templates" (electronic forms filled in with standard set configurations) available on an administrative terminal with a CRT display, equipment number programming was speeded up.

The obvious alternative was to relate user features to the prime line directory number of a set, so that only the directory number to equipment number translation would have to be changed when a user moved from one office to another; the prime line would bring all its programming with it to the new matrix port. An even easier approach was developed, as exemplified by Northern Telecom's "set relocation feature." Here, the user about to move picks up the phone, keys in an "I'm moving" code and unplugs the phone. At the new desk, the phone is be plugged in (assuming the line from the wall jack goes to the right kind of line card at the PBX), the user keys a "Here I am" code and the prime line number, and the job is done.

Going one step further, systems such as the SRX and the late TeleNova include a unique serial number burned into a ROM in their proprietary telephone sets. A telephone SET is assigned to a user, and when the phone is plugged in, the system reads the serial number before doing anything else. Then, using a look-up table that relates set serial number to prime line number (it already knows the equipment number), it makes the set with all its features and privileges available for use. All the user did was plug the set in at its new location. Stories are told of users who carry their phones with them, even to the company lunch room, and plug them in at any convenient phone jack to get their calls. AT&T uses a similar serial number to allow a system control to identify circuit boards when installed so that it can prevent a system administrator from such mistakes as assigning an analog phone to a digital line port.

Clearly, allowing the system to identify a phone as soon as it is plugged in is only a jump away from one of the main capabilities of micro-cell wireless telephone systems presently being discussed. With these systems, the each user has a personal telephone with a unique number, but there is no corresponding physical port on the matrix; rather, users enter the system on whatever radio channel is available to the closest of several shared radio transceivers. Originating calls is no problem; the set sends its identity along with the called number. Receiving a call, however, is harder. Unlike wired systems, where the location of the set is known once the station relocation procedure is completed, portable radio phones don't necessarily check in as their owner wanders around. Thus the system has to search for them via its several base stations, locating them, at least momentarily, when they respond. The point is, however, that a numbering plan for mobile radio phones has only the set on which to hang an identity and class of service; the equivalent of an equipment number can change from one call to the next, and even during the time a call is in progress if the caller is walking around.

It is interesting to note that much of the activity in micro-cell radio phones (PCN, PCS, etc.) is copying the mistakes of PBX manufacturers in the 1970s. That is, single line phones dominate the literature. Exactly how the 1A2 key functions will be handled with PCNs is not clear; what is clear is that without these basic functions, PCNs will be little more than a novelty filling a useful but limited niche in business telecommunication.

ISDN And Voice/Data

Adding a data channel to a digital phone, as has been done with most proprietary digital PBX phones and is intended for ISDN phones for both PBX and Central Office switches brings us a new array of problems. Now one "port" on a PBX line card has to be able to handle both voice and data simultaneously on a circuit-switched basis using the B or B-like channels, and packets on the D or signaling channel. Other functions, such as video teleconferencing, require two channels at present, and in the future, various data functions may require three or more channels combined into one to get the bandwidth required for certain kinds of transmission. To add to the fun, some versions of ISDN talk about having up to 8 devices on any given line, sharing the (present) two B channels and the D channel. Although only two devices can be reached by circuit switching at the same time via the B channels, all the rest are available simultaneously, at least in principle, to packets on the D channel. Now the numbering plan will have to include device, and the switch will have to select the proper transmission path via a physical port and line.

Conclusions

The time has come to carefully rethink the concept of PBX numbering plans. Between the portability implied by automatic set relocation and micro-cell radio phones, as well as multiple terminal devices per ISDN port, it appears that the basic entity in a communication system must be the communication terminal itself (phone, computer, or whatever). However, that terminal, particularly in a business context, will usually not act alone. It will require some outside controlling force to provide support and to associate it with other terminals for secretarial screening, peer-group interaction, access to voice mail and file servers, etc. These relationships, particularly in a mobile environment, will have to be continuously redefined in real time.

Already we have seen examples of cellular radio phones being rung in parallel with fixed phones, allowing the called party to answer in either car or office. And new approaches to paging allow an incoming call to be held until the called party can be located and alerted; common channel signaling then allows the called party to respond from anywhere and be connected to the caller. If the called party cannot be found, the caller can be connected to a voice mail system to leave a message. But we are just beginning to explore some of these possibilities. It is important that we not become so carried away with the new that we forget existing features which experience proves we need. But however it goes, we will still have to relate each phone, its current location, and the information about it that the rest of the system needs for operation, administration, billing and other functions.

If we are going to have a variety of competing companies, world wide, all providing systems, we are going to have to go far beyond anything we have done so far in standardizing numbering plans and methods of operation. I, for one, am giving serious thought to breeding carrier pigeons.

Sidebar. Party Lines and Numbering Plans

Not too many years ago, party lines were common, providing economy service to several customers sharing one line and one port on the CO switch. Usually some form of distinctive ringing had to be used to get the right party to answer, but more important, callers had to have a way to identify the particular party they wanted. One way was to add a party-line letter such as W, R, M or J. The caller would ask for 2314W, the operator would plug into the jack labeled 2314 and push the ringing button associated for stations designated W. In SXS systems, party line connectors had a built-in auxiliary switch to take the last digit, dialed after the call was set up, to select the ringing.

Systems of this sort were called "terminal per line." To obtain more uniform numbers without appended letter designations, "terminal per station" operation was developed. Here, the same line might be terminated on connectors in two different line groups, each providing a different kind of ringing. A caller would dial one number to reach a party on one line group, and a different number to reach another party on the same pair of wires but accessed via the other line group.

With the coming of crossbar systems, trunk circuits were able to apply many different kinds of ringing to one line based on translation of the number dialed by the caller. Here the best of both words was obtained: the user didn't need to dial a party line letter, and the telephone company did not have to waste terminals on the switching matrix.

With party lines thus masked, the problem of "reverting calls," where one party calls another on the same line, became more difficult. Somehow, the caller had to be made to hang up so that the party line could be rung; signals were provided for both parties to hear, and when the calling party heard ringing stop, the call had been answered and it was time to pick up the phone before the called party, finding nobody there, abandoned.

AT&T's ESS field trial in Morris, IL, back in the 1960s, used tone ringing. Eight different audible tones were available, and once the caller had been convinced to hang up, it was easy to apply alternately the ringing needed by calling and called. This feature was quickly adapted to an intercom function for homes with several telephones. By providing a different ringing frequency for each phone, and using a special intercom code, Mother in the kitchen could call Dad in the living room by dialing an intercom number, hanging up, and waiting for ringing to stop.

A somewhat different version of specialized ringing is being marketed today to allow residential customers to know which family member is being called, or who is placing an incoming call. One line can be reached by a several different directory numbers, and each directory number will cause a different kind of ringing to be applied. Two short rings might suggest a business call for Dad, whose associates dial him with 555-1234, while a short and a long might alert Junior, whose friends are instructed to dial 555-3421. Mother might tell her friends her phone number is 555-7442, and the system would ring with two longs. This approach certainly gives the telephone company a service it can sell but, like Centrex, it isn't doing much to help solve the telephone number shortage.

[ Top ] [ Next ] [ Table of Contents ]

Copyright 2006 Lee Goeller. All Rights Reserved.