Talk:IBM 2741

This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects:

Computing Mid‑importance This article is within the scope of WikiProject Computing, a collaborative effort to improve the coverage of computers, computing, and information technology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.ComputingWikipedia:WikiProject ComputingTemplate:WikiProject ComputingComputing articles Mid This article has been rated as Mid-importance on the project's importance scale. This article is supported by Computer hardware task force (assessed as Mid-importance).

The 2741 originated at the Thomas J. Watson Research Center. A project led by Steven Firth created the 2741 (and drove it to market) to support the ATS/360 ("Administrative Terminal System") software product, a text editing and storage system which otherwise would have had no commercially practical user terminal. ATS/360 was implemented in IBM/360 assembler; it first ran under DOS and then was migrated to MFT. ATS/360 included its own proprietary time-sharing software as did other 360 time sharing applications of those days. I know because I extensively modified and extended ATS/360 in the late 1960s and early 1970s.

I first remember the 2741 from CRBE, and later with WYLBUR. It wasn't until the 3705 that I could use ASCII terminals with WYLBUR. Gah4 (talk) 01:43, 19 October 2016 (UTC)[reply]

I know that APL\360 supported the 2741 because I wrote the code. There was a DOS system for text-editing which I believe supported the 2741. I cannot remember its name. It is also likely that IMS, GIS and CICS supported the 2741. I cannot verivy this.Rdmoore6 21:54, 3 January 2006 (UTC)[reply]

A few years earlier (1968 or so) I saw a time sharing system offering Fortran capabilities on a 2741. We used the Manifold element with it. I don't know what the host OS or computer was, though. Jeh 07:16, 13 September 2007 (UTC)[reply]

In case it is of any interest, from 1972 thru 1975, I wrote CAI modules [subject: organic chemistry nomenclature] using a 2741 connected via a modem and leased line to an IBM 360/155 located at the SUNY Binghamton campus sixty miles west of the CCC campus. I still have that APL typeball somewhere in a box packed away from those days. And wonderful days they were. I should never have left software to go into electrical engineering I can tell you!  :-) JimScott 07:56, 30 September 2007 (UTC)[reply]

In the image of the 2741 APL keyboard layout used in this article, I think it's missing an Greek IOTA over the "I". Shift-"I" should give an IOTA for APL. --- Wikiklrsc (talk) 02:26, 6 February 2010 (UTC)[reply]

That's what it says in my refs too. Jeh (talk) 05:32, 6 February 2010 (UTC)[reply]

Thanks, Jeh. That's the way it was when I used APL in the 1960s. Maybe it's a trick? Like the cartographer's trick: leave out some town or put one fictitious town in to prove copyright? Anyway, I have no easy idea how to get the IOTA back on the keyboard layout in the image. Bests. --- (Bob) Wikiklrsc (talk) 18:40, 6 February 2010 (UTC)[reply]

I wrote to the uploader and he uploaded the correct keyboard layout. Bests. --- (Bob) Wikiklrsc (talk) 18:58, 6 February 2010 (UTC)[reply]

Google image search provides three hits for "IBM 2741". Two of these hits are now external references from the article. All three Google hits seem to have different scanned versions of the same photograph. I suspect the photo came from an IBM manual. I asked the person who maintains the acis/history/2741.html page at Columbia. is replay was: "It's not my photo. I have no idea whose it is." It would be nice to have the photo on wiki (columbia U has best scan) if permission could be obtained from IBM or whoever took the photo.Rdmoore6 21:54, 3 January 2006 (UTC)[reply]

There are more photos of what I believe are IBM 2741's built into BCL SUSIE machines at http://www.computinghistory.org.uk/det/9515/BCL-Susie/ P.r.newman (talk) 10:59, 10 September 2011 (UTC)[reply]

Not the whole of a 2741 but almost certainly the mech. Hope the article text is a suitable compromise. No reason to not link the page either. Jeh (talk) 21:33, 11 September 2011 (UTC)[reply]

There is a question about IBM 2741's coming in non-desk versions. I do remember the portable versions, but don't remember looking at the sticker to see who made it. I only remember the white desk like the bottom picture, not the wood grain desk like the APL picture. I had thought that was a portable version sitting on a desk, but it is hard to tell from the picture. I used these a lot 40 years ago, and even was allowed to bring a portable one home one night. (There weren't many for that use.) I do remember the portable version had more electronics in the back, that the APL picture doesn't show. What were the clones called? Gah4 (talk) 01:37, 19 October 2016 (UTC)[reply]

In the picture of the wood grain desk you'll notice a black "rim" around what appears to be the bottom of the machine. This is the same rubber bushing that is obvious when you see the usual 2741 embedded into the square white desk. I notice foreign-language documents on the wood grain desk - maybe that desk was a "special" sold in another country. Or, I suppose it is possible that somebody took a 2741, removed the Selectric mechanism from the desk, and put the electronics package (which was not small!) under the desk of their choice. But all of the pictures in the "IBM 2741" documents do show it built into a square white desk. Jeh (talk) 08:31, 19 October 2016 (UTC)[reply]

Anyway, it was that desk that fooled me. But then, that one doesn't have the extension on the back that I, sort of, remember for the non-desk (clone) versions. Very interesting picture! Gah4 (talk) 09:36, 19 October 2016 (UTC)[reply]

My guess is that the "foreign language" is Icelandic, since it contains "eth"s (wikipedia:Icelandic orthography). The 2741 is pure IBM with an APL keyboard. Peter Flass (talk) 23:20, 21 October 2016 (UTC)[reply]

Did IBM make the wood grain desks? As above, someone might have replaced the IBM desk, moving all the electronics and mechanics. (And, I suspect, terminating any warranty coverage.) I think all the APL stories I ever knew, people put APL balls on regular 2741s. (So the wrong keyboard.) Maybe real APL 2741s have wood grain desks? Very strange. Gah4 (talk) 23:38, 21 October 2016 (UTC)[reply]

Does this really belong here? It's more about the Selectric and its interchangeable elements than about the 2741. Those Algol 68 documents might as well have been prepped on a Selectric with the "correspondence coded" APL element. There is no mention or evidence here that an actual 2741 terminal was used. I think this deserves to be a sentence or two in the IBM Selectric typewriter article (it already mentions the prevalence of the IBM Symbol element's output in various math and science papers). Jeh (talk) 19:23, 11 December 2016 (UTC)[reply]

Someone inserted, "Electrically, it is related to the Break key on ASCII terminals.". I don't know what that means, and there is no explanation, nor a citation. In fact, I do not believe it is even true. Unlike ASCII terminals, the 2741 keyboard is mechanically locked so that no keys can be depressed while the remote end is controlling. Pressing the ATTN key unlocks the keyboard. I used the 2741 for ATS and later, APL, for years, and that's not how any of the hundreds of ASCII terminals that I ever used worked. I have looked over the IBM schematics for the 2741 and I do not see how the circuits are the same as for an ASCII terminal, nor the line protocol; I just don't know why someone wrote that these are the same things. I suspect it is just made-up. Perhaps the editor meant to suggest that these two keys were sometimes used for similar purposes in some ways. I would like to see some evidence that these are "Electrically related". Or else the comment should be removed.Dicirnah (talk) 00:50, 26 April 2017 (UTC)[reply]

I can tell you one thing for certain, and that is that the break key on an ASCII terminal doesn't send a character as such. Instead it generates a framing error: A start bit that is not followed by a stop bit in the expected time; that is, the line is held at "spacing condition" for longer than a character time. The receiving system detects a "break" by noting that a "stop bit", marking condition, has not shown up at the required time after the "start bit".

The rest of this is partly speculation, recollection from decades-old memories (no doubt partly faulty), etc....

Looking at the 2741 code set, we see that there is no character for "ATTN". I believe it actually did generate a framing error (it pretty much had to do one or the other: generate a character, or a framing error) but I have no way to test that now. Note that generating a framing error is easy: Just force the RS232 output to "spacing" condition for more than a full character time. I believe most ASCII terminals settled on two or three character times. The ASR33 would just hold the line at "spacing" for as long as you pressed the key. Before computers this was used on TTY-to-TTY links to get the other operator's attention. The receiving TTY didn't really care about framing errors but the continuous "spacing" condition would make its printer "cycle" continuously without printing anything.

If you can trace the schematic/logic diagram backwards, from the TxD line to the input of the RS232 level shifter that feeds the TxD line back to the UART serial output (mind you, the UART was built with random logic), you may find a gate in there that overrides the UART output and forces the line to "spacing" regardless of what the UART is doing.

Note that the ATTN key cannot unlock the local machine by itself. I once worked on a very slow APL system under which the ATTN key only produced results s-l-o-w-l-y when the system was loaded. However the ATTN keypress is sent, the result is that it causes the remote system to stop what it's doing and send a circle-C, "end of message" character. Receipt of the circle-C is what actually unlocks your keyboard... if I recall correctly.

If you can give me a link to the 2741 schematics online I could look also. I bought the set years decades ago, intending to do my senior project in that space, but we ended up with no 2741-compatible system to connect to so I had to shift to something else. Jeh (talk) 02:54, 26 April 2017 (UTC)[reply]

Never mind. Found it here: http://bitsavers.informatik.uni-stuttgart.de/pdf/ibm/27xx/2741/GA24-3415-3_2741_Data_Terminal_Aug72.pdf And no need to pore over IBM's line-printer-printed logic diagrams.

From page 13:

The operator presses the Attention key to cause an interrupt. This causes the 2741 to transmit a 200 to 360ms continuous space signal. This signal requests that the computer program end transmission by transmitting a ©. The program mayor may not honor this request. If transmitted by the CPU, the © places the 2741 in transmit mode. The terminal then transmits a (circle-D) and its keyboard unlocks.

So, yes, electrically, the ATTN key is very much like the BREAK key. they both generate a spacing condition on the line, resulting in the host seeing a framing error. Note that even the shortest period mentioned, 200 ms, would be a framing error at 134.5 bits/sec since it would be about 27 bit times, almost three full character times, far longer than the time before another stop bit should have shown up.

However the result of the remote host receiving a framing error was not necessarily the same. In a system hosting a 2741, the host should stop sending whatever it was sending, set up to expect data from the 2741 (often by terminating the current program and, depending on th te system, sending a command prompt), and send a circle-C to unlock the 2741's keyboard.

But ASCII terminals had no "lock the keyboard" concept at all, let alone a character code that would unlock them, and their host systems' response to a framing error ("break" key) was much more varied. In some systems, yes ,the "break" signal simply terminated the current program. But in at least one timesharing system I worked on the BASIC dialect had an ON BREAK statement that would allow the program to take a branch if break was received. It might even continue sending stuff to the terminal. In another system the FORTRAN runtime system had a function you could call to find out if "break" had been seen, and then you could take a conditional branch depending on that result.

So in that regard, you're right: yes, the ATTN and BREAK keys could be very different in their effect. Certainly a "break" key on an ASCII terminal never resulted in a keyboard being unlocked!

Thanks for asking! Jeh (talk) 03:40, 26 April 2017 (UTC)[reply]

Thanks for your contribution and edits! (Although I did chuckle when you said "UART" which pedantically didn't exist back then.) Good work with cite. Dicirnah (talk) 19:49, 26 April 2017 (UTC)[reply]

Which is why I wrote " (mind you, the UART was built with random logic)" right after that. Again, I did have the service manual including the logic diagrams on the 2741 at one time... so I know there was not a WD1402A in it. ;) But the function obviously existed. Just btw, do you have any recollection of what a 1402A or a Signetics 2651 cost in, say, the early 80s? (Asking for a friend.) Somehow I do seem to have misplaced my Digi-Key, etc., catalogs from that era. Jeh (talk) 21:04, 26 April 2017 (UTC)[reply]

With 2741's and ASCII terminals attached to a 3705, more specifically, the way they are used with WYLBUR, BREAK on the ASCII terminal has the same function that ATTN does on the 2741. As far as I remember, the 3705 ignores any characters coming in while it is sending, but does recognize BREAK. On the other hand, there are some ASCII systems that also work with 2741s. I believe that some DEC systems do, and for them ATTN would have to be similar to the control-C or control-Y from ASCII terminals. Oh, also, you can use ATTN even when the host isn't sending. Gah4 (talk) 08:11, 26 April 2017 (UTC)[reply]

Why would ATTN have to be similar to ctrl-C or ctrl-Y? DEC's serial line adapters and muxes can recognize framing errors (ie BREAKs) just fine. Then the DEC system can stop what it's doing and send the required circle-C. SET TERMINAL/DEVICE_TYPE=2741 even existed on VMS, though not in current versions. Jeh (talk) 08:46, 26 April 2017 (UTC)[reply]

Similar in that is how you stop the system from doing what it is doing, and get control back again. If a program is in an infinite loop, on an ASCII terminal you can type control-C, but the only thing you can do on a 2741 is ATTN. I don't know if VMS would consider it like ctrl-C or ctrl-Y, though. Gah4 (talk) 18:33, 26 April 2017 (UTC)[reply]

Not from an ASCII terminal, it won't. Might be an interesting bit of research. It's possible that the 2741 was only supported in a limited way, say as an output device. I never heard of a 2741 connected to a VMS system, let alone saw one. Just noticed the "134" bit rate setting and the device type option. Jeh (talk) 18:45, 26 April 2017 (UTC)[reply]

I do remember DECwriters as spooled line printers, so I suppose you could also do that with a 2741. My time with 2741s didn't overlap my time with VMS, so I never got to try them together. Gah4 (talk) 18:54, 26 April 2017 (UTC)[reply]

I was going to disagree with the removal of the mention of control-C, but I suspect that the reference will be lost to enough readers, that it isn't so useful. Though Control-C#In_command-line_environments might not be so bad. Gah4 (talk) 08:18, 26 April 2017 (UTC)[reply]