jjacobs@well.UUCP (Jeffrey Jacobs) (01/27/88)
In <5174X@utah-cs.UUCP>, my good friend Stan Shebs writes: >Jacobs must be angling for DoD money or something. I don't think I've ever >heard anybody, not in the pay of the DoD, say anything good about Ada... >Tony Hoare's Turing lecture certainly had some critical remarks, to the >effect that we risk missiles hitting our own cities by using Ada. >(I don't know what he thinks about CL.) As usual, Stan ignores what I say in favor of his own strawman. The process and thought that went into designing Ada resulted in a much better language than CL. Ada has much better import/export/package facilities, better typing facilities and a much more coherent syntax. I am talking about *quality* of the final language. As to missiles hitting our cities, I often think that the best hope for survival of the human race is the fact that a lot of this stuff probably won't work if it's ever really used. >>It should >>be embarassing to everybody in the field that most shells and tools are >>no longer written in LISP. >It *is* embarassing, but CL is not the reason; the same thing would have >happened if (for instance) Scheme had been standardized on.But enough of this >random flaming; let's get to the meat of one of my arguments. I disagree; I also cover this topic in the upcoming March issue of AI Expert in an article entitled "Quo Vadis, Lisp?" We can cover this when that appears. >>CL is a nightmare; it has effectively killed LISP development in this >>country. >You're going to need some facts to back up that assertion. I see plenty >of Lisp work going on. Ok! Let's look at the facts; I hope that some of the rest of the net will contribute information on whatever they are doing. Let's also keep "Lisp work" confined to development and implementation of Common Lisp, *not* applications, editors or what have you. I would appreciate any corrections or contradictions to the following. Rumors are clearly identified as such. Commercial: 1 There are only two CL vendors selling CL implementations for multiple standard architectures, Lucid and Franz. Estimated number of copies sold by each is ~2,000, for a total of 4,000. 2. One major vendor for MS-DOS, Gold Hill. Number of copies sold unknown, but of the (nearly) Steele Complete 286/386 versions, probably under 3,500. Several minor vendors selling incomplete subsets. Franz/Coral only true Common Lisp for Macs. Out of the three mentioned above, only Franz is profitable; Gold Hill is (rumored) to be close to self supporting, depending on how you look at it. 3. DEC continues development of it's own VAX-Lisp; virtually all other hardware vendors have abandoned or are planning on abandoning internal development in favor of OEM'ing from Franz or Lucid. This includes HP (Rumor). 4. Lisp Machines Inc. is history. 5. Symbolics is in deep yogurt, losing money, laying off people, unable to issue more stock and firing its Chief Financial Officer. Rumor has it that they are madly trying to develop a 68020 version. 6. Xerox Common Lisp. MIA (Missing In Action). Apparently still not ready to release. **Rumor** has it that Xerox will abandon CL running on their own hardware and OEM Sun Sparc architecture running one of the above. (Repeat; **rumor**). Universities: 1. NIL, featured prominently on the cover of Steele, was stillborn. MIT, the birthplace of Lisp, unable to produce a Common Lisp. 2. CMU is the only university to produce an effectively Steele-complete version of CL, aka Spice. (I don't know how complete Hedrick's DEC-20 Common Lisp is. Charles?) 3. University of Utah, as of the last time Stan and I went round and round, still only had a subset of Common Lisp. By Stan's own words, an incomplete implementation is "broken" :-). Are there any Universities seriously working on full scale, Steele-complete CL implementations? If so, let's hear about them. (There should be adequate funding guaranteed so that these implementations stand a reasonable chance of being "Steele-complete"). >I place the blame on lazy and timid Lisp implementors who forgo >optimizations because "they would compromise Lisp tradition", and >companies who get away with >selling shoddy systems because there is little or no competitition. Gee, whatever happened to blaming it on the failure of hardware to keep pace? Of course, we are all awaiting Stan's version that will blow all the lazy implementors away, and sweep the world. Did you ever stop to consider that there is so little competition because there is so little demand? Jeffrey M. Jacobs CONSART Systems Inc. Technical and Managerial Consultants P.O. Box 3016, Manhattan Beach, CA 90266 (213)376-3802 CIS:75076,2603 BIX:jeffjacobs USENET: jjacobs@well.UUCP
kers@otter.hple.hp.com (Christopher Dollin) (01/27/88)
/ otter:comp.lang.lisp / jjacobs@well.UUCP (Jeffrey Jacobs) / 4:39 am Jan 27, 1988 / In <5174X@utah-cs.UUCP>, my good friend Stan Shebs writes: >Jacobs must be angling for DoD money or something. I don't think I've ever >heard anybody, not in the pay of the DoD, say anything good about Ada... >Tony Hoare's Turing lecture certainly had some critical remarks, to the >effect that we risk missiles hitting our own cities by using Ada. >(I don't know what he thinks about CL.) As usual, Stan ignores what I say in favor of his own strawman. The process and thought that went into designing Ada resulted in a much better language than CL. Ada has much better import/export/package facilities, better typing facilities and a much more coherent syntax. I am talking about *quality* of the final language. As to missiles hitting our cities, I often think that the best hope for survival of the human race is the fact that a lot of this stuff probably won't work if it's ever really used. >>It should >>be embarassing to everybody in the field that most shells and tools are >>no longer written in LISP. >It *is* embarassing, but CL is not the reason; the same thing would have >happened if (for instance) Scheme had been standardized on.But enough of this >random flaming; let's get to the meat of one of my arguments. I disagree; I also cover this topic in the upcoming March issue of AI Expert in an article entitled "Quo Vadis, Lisp?" We can cover this when that appears. >>CL is a nightmare; it has effectively killed LISP development in this >>country. >You're going to need some facts to back up that assertion. I see plenty >of Lisp work going on. Ok! Let's look at the facts; I hope that some of the rest of the net will contribute information on whatever they are doing. Let's also keep "Lisp work" confined to development and implementation of Common Lisp, *not* applications, editors or what have you. I would appreciate any corrections or contradictions to the following. Rumors are clearly identified as such. Commercial: 1 There are only two CL vendors selling CL implementations for multiple standard architectures, Lucid and Franz. Estimated number of copies sold by each is ~2,000, for a total of 4,000. 2. One major vendor for MS-DOS, Gold Hill. Number of copies sold unknown, but of the (nearly) Steele Complete 286/386 versions, probably under 3,500. Several minor vendors selling incomplete subsets. Franz/Coral only true Common Lisp for Macs. Out of the three mentioned above, only Franz is profitable; Gold Hill is (rumored) to be close to self supporting, depending on how you look at it. 3. DEC continues development of it's own VAX-Lisp; virtually all other hardware vendors have abandoned or are planning on abandoning internal development in favor of OEM'ing from Franz or Lucid. This includes HP (Rumor). 4. Lisp Machines Inc. is history. 5. Symbolics is in deep yogurt, losing money, laying off people, unable to issue more stock and firing its Chief Financial Officer. Rumor has it that they are madly trying to develop a 68020 version. 6. Xerox Common Lisp. MIA (Missing In Action). Apparently still not ready to release. **Rumor** has it that Xerox will abandon CL running on their own hardware and OEM Sun Sparc architecture running one of the above. (Repeat; **rumor**). Universities: 1. NIL, featured prominently on the cover of Steele, was stillborn. MIT, the birthplace of Lisp, unable to produce a Common Lisp. 2. CMU is the only university to produce an effectively Steele-complete version of CL, aka Spice. (I don't know how complete Hedrick's DEC-20 Common Lisp is. Charles?) 3. University of Utah, as of the last time Stan and I went round and round, still only had a subset of Common Lisp. By Stan's own words, an incomplete implementation is "broken" :-). Are there any Universities seriously working on full scale, Steele-complete CL implementations? If so, let's hear about them. (There should be adequate funding guaranteed so that these implementations stand a reasonable chance of being "Steele-complete"). >I place the blame on lazy and timid Lisp implementors who forgo >optimizations because "they would compromise Lisp tradition", and >companies who get away with >selling shoddy systems because there is little or no competitition. Gee, whatever happened to blaming it on the failure of hardware to keep pace? Of course, we are all awaiting Stan's version that will blow all the lazy implementors away, and sweep the world. Did you ever stop to consider that there is so little competition because there is so little demand? Jeffrey M. Jacobs CONSART Systems Inc. Technical and Managerial Consultants P.O. Box 3016, Manhattan Beach, CA 90266 (213)376-3802 CIS:75076,2603 BIX:jeffjacobs USENET: jjacobs@well.UUCP ----------
kers@otter.hple.hp.com (Christopher Dollin) (01/27/88)
Re: University implementations of Common Lisp ... Sussex University in Britain include Common Lisp as part of the Poplog environment. Although Poplog is SOLD by Systems Designers, it is BUILT at Sussex. Poplog V13 includes a pretty damn near complete implementation of VL, sorry, CL. Not bad going for one main implementor plus the Poplog VM genius in a couple of years ... Poplog includes an integrated editor and the core language (Pop11) with Prolog and Common Lisp. All these languages can talk to each other. Standard ML is also available. Regards, Kers | "Why Lisp if you can talk Poperly?"
brad@pheasant.cs.utexas.edu (blumenthal @ home with the armadillos) (01/28/88)
[Gee, you guys are fun to watch.] >>2. CMU is the only university to produce an effectively Steele-complete >>version of CL, aka Spice. (I don't know how complete Hedrick's DEC-20 >>Are there any Universities seriously working on full scale, >>Steele-complete CL implementations? > >UMass Amherst has got a distributed Common Lisp project going, and I think >Rochester has something underway for the Butterfly. Can anyone say "Kyoto?" Complete, fairly portable, and you can't beat the price. Take care, brad Brad Blumenthal ARPA: brad@pheasant.cs.utexas.edu UUCP: {seismo, harvard}!ut-sally!pheasant!brad
hedrick@athos.rutgers.edu (Charles Hedrick) (01/28/88)
>2. CMU is the only university to produce an effectively Steele-complete >version of CL, aka Spice. (I don't know how complete Hedrick's DEC-20 >Common Lisp is. Charles?) DEC-20 Common Lisp is a port of CMU's Spice. It is as complete as Spice was at the time it was done. The part we did is complete, but that is low-level functions and the compiler. I was never able to be sure how complete the stuff we took from CMU was, as we were depending upon the CL validation suite to do final checkout. As far as I can tell, it never materialized. Had the DEC-20 survived, we would have kept manpower on the project and made sure it was complete, but it didn't seem worth doing. (DEC-20 CL was really built for the Jupiter. There was at least one design tradeoff made on the basis of preliminary instruction timings, and it was built to take advantage of the bigger virtual address space simply by changing some contants. Once the Jupiter was cancelled, our enthusiasm waned rapidly. In my view, the existing model isn't quite powerful enough for CL. Typical configurations just don't have enough memory to deal with a number of users running CL, though one or two people can run it without trouble. I understand it was actually used for coursework at Stanford. We never had that much courage.) Certainly all the low-level primitives, compiler technology, etc., is there, so it's just a matter of updating the Lisp system code to the newest version of the CMU stuff. It is quite true that a complete CL is beyond the ability of the typical university hacker to produce. However if you start with Spice Lisp, producing a CL is probably not much worse than any other Lisp dialect. It's certainly within the realm of a team with a few good people in it. It doesn't need anything like the resources that were used to produce some recent big systems, e.g. X. However if you had to do it from scratch (i.e. without Spice Lisp), and you had to do it within a couple of years (Spice Lisp took far longer to produce something complete - which is why so many CL's based on Spice Lisp started out as half-baked), it would probably require a team that only a couple of big-name institutions could put together. It's a fact of life that software is getting bigger. It's hard for one guy to produce a major package these days (except if he's RMS).
ok@quintus.UUCP (Richard A. O'Keefe) (01/28/88)
In article <5084@well.UUCP>, jjacobs@well.UUCP (Jeffrey Jacobs) writes: > Commercial: > > 1 There are only two CL vendors selling CL implementations for multiple standard > architectures, Lucid and Franz. Estimated number of copies sold by each > is ~2,000, for a total of 4,000. > There is a system called PopLog which was produced at the University of Sussex, in the UK. This is is Pop11 + Prolog + CL, all intercallable. I have not seen a recent version, so can't comment on the completeness of their Common Lisp, but I expect good things from them (for example, their Prolog was slow, but very close to the de-facto "DEC-10" standard). PopLog has oodles of on-line help and source-code libraries. Systems Designers Ltd market it commercially in the UK; they have a US distributor but I don't remember the name. Last I heard they had sold a couple of hundred copies, but that was a few years ago. There were VAX/VMS, VAX/Unix, various M680x0, and SUN versions being distributed, and other versions were in the works. > 6. Xerox Common Lisp. MIA (Missing In Action). Apparently still > not ready to release. **Rumor** has it that Xerox will abandon > CL running on their own hardware and OEM Sun Sparc architecture > running one of the above. (Repeat; **rumor**). > XCL is not missing in action. I've got it. Works fine, as far as I can tell. Ask for the Lyric release. I haven't tried everything in The Book, but everything I've tried was there, and they provide the "proposed" error handling system (errors are objects). The rumour I've heard is that Xerox will port THEIR OWN software to some standard hardware (which needn't entail abandoning their own hardware) I have not heard this myself from anyone at Xerox, neither have I asked. I could tell you about some other rumours I've heard, but what's the point? They're probably all wrong. I'm not sure that the number of Common Lisp vendors (whether Mr Jacobs has the right number or not) tells us much about what Common Lisp has done to the Lisp market. How many commercial vendors were there BEFORE? Surely the troubles of LMI and Symbolics cannot be blamed on Common Lisp; CL is too similar to ZetaLisp for that!
fritz@hpfclp.HP.COM (Gary Fritz) (01/29/88)
> >3. DEC continues development of it's [sic] own VAX-Lisp; virtually all > >other hardware vendors have abandoned or are planning on abandoning internal > >development in favor of OEM'ing from Franz or Lucid. This includes > >HP (Rumor). > > Not a rumor; it's true. The hardware vendors fell into the trap of assuming > that good software is easy to build - a common misconception of hardware > types who "once wrote a 200-line Fortran program". In fact, hardware vendors > have *rarely* succeeded in producing good software of *any* sort, but it > doesn't ever seem to stop them from trying... This is an unfair statement. HP Common Lisp was developed by a group of extremely qualified software engineers, many with extensive previous experience in building Lisps. Granted, the implementation had some problems, but those were due more to historical/etc. reasons than to "hardware types who once wrote a 200 line FORTRAN program". Just because someone works for a company that produces hardware, that doesn't make him/her a software illiterate. "Hardware" companies hire software types too, ya know. JJacobs, there are reasons for "abandoning" proprietary CL's other than a supposedly brain-damaged lanaguge spec. As an example, there were *no* acceptable CL's to pick up when HP began developing its own. However, there are now some obvious choices of "de facto" standard CL's (such as Lucid), and it doesn't make sense to expend engineering effort on maintaining a non-mainstream language. Lucid has many engineers devoted to the care, feeding, and improvement of their CL, and it makes sense to join their effort. Gary Fritz
welch@tut.cis.ohio-state.edu (Arun Welch) (01/29/88)
>6. Xerox Common Lisp. MIA (Missing In Action). Apparently still >not ready to release. **Rumor** has it that Xerox will abandon >CL running on their own hardware and OEM Sun Sparc architecture >running one of the above. (Repeat; **rumor**). MIA? Xerox Common Lisp was released in August. By now, it's even been released in Europe to Rank Xerox sites. Xerox and Sun announced in October that they are moving all Xerox products off of Xerox's custom hardware to Suns SPARC architecture. This includes XDE, the XNS services, Xerox Lisp, Viewpoint, and so on. This information was rather widely publicised, it's not a rumor. >1. NIL, featured prominently on the cover of Steele, was stillborn. MIT, >the birthplace of Lisp, unable to produce a Common Lisp. I don't know what you mean by stillborn. I used it for about 2 years, and I know of other people in other parts of the country who did too, both at educational institutions as well as commercial sites. >2. CMU is the only university to produce an effectively Steele-complete >version of CL, aka Spice. (I don't know how complete Hedrick's DEC-20 >Common Lisp is. Charles?) I've used Hedrick's DEC-20 CL, and haven't found any defficiencies. If it's missing something, it's not in the documentation as being missing, nor is it a major enough thing to be critical. Amazingly enough, you're also missing a rather large contender in the field: TI. They're doing quite well selling their Explorrer II's and LX's, and are rumored to be about to announce a plug-in board for the Mac II which is about half way between the Explorer I and the Explorer II in performance. Not to mention that their lisp chip is gonna have all kinds of interesting applications in other parts of their organisation. In fact, last I heard, Explorer II's were back-ordered a couple of months. To find TI missing from your list is pretty amazing. Or is it because you can't find anything bad to say about them? Yet another vendor missing from your list is HP. I haven't used HP's CL, so I can't comment on it. And another vendor out there is Ibuki, who are marketing a supported, enhanced version of KCL. Another case of my not having used it enough to comment on it. And another vendor missing from your list is BBN ACI, who report quite good sales of their Butterfly SCheme, and will probably report even better sales of Butterfly Lisp (which is being developed on HP Bobcats, so HP must have a pretty good Lisp...). Yes, BBN ACI recently re-organised, but that was to accomodate expansion, not retrenching for a loss of market. Butterfly Lisp is waiting on Mach to come out on the Butterfly, so it's not expected out until June. Lisp development seems to be going along pretty well in a number of organisations. How else can you explain the emergence of CLOS and CLX, one as an actual standard, and the other which will probably develop into a de facto standard, with people already writing toolkit's for both of them (as evidenced by TI's CLUE and Xerox's PCL-Environment). There seems to be a lot missing from your flame, not to mention some disinformation... I hope your article in AI Expert doesn't reflect more of the same... ...arun ----------------------------------------------------------------------- Arun Welch Lisp Systems Programmer, Lab For AI Research, Ohio State University welch@tut.cis.ohio-state.edu
darrelj@exec.sm.unisys.com (Darrel VanBuer) (01/29/88)
In article <5084@well.UUCP> jjacobs@well.UUCP (Jeffrey Jacobs) writes: >In <5174X@utah-cs.UUCP>, my good friend Stan Shebs writes: > >6. Xerox Common Lisp. MIA (Missing In Action). Apparently still >not ready to release. **Rumor** has it that Xerox will abandon >CL running on their own hardware and OEM Sun Sparc architecture >running one of the above. (Repeat; **rumor**). > Xerox Common Lisp is out, and has been for a few months. Deliveries have been slow because of the magnitude of changes, and Xerox has apparently trying to minimize the number of customers on the steep part of the learning curve at one time. Beta testing went faster than planned because of the relative solidity of the job. The main weakness of what was delivered in the first release is that many of the Interlisp tools don't grok common lisp yet (e.g. Masterscope). It does seem that Sun and Xerox have agreed in principle to migrate Xerox software to Sun hardware (with Sun committed to support for XNS), but indications are in products for a couple of years out. I guess Xerox finally decided their application customer base is too small to support competitive hardware development (the Star/Viewpoint text processing software is also migrating. A year was a long time to be overdue, though.
mincy@think.COM (Jeffrey Mincy) (01/29/88)
In article <675@athos.rutgers.edu> hedrick@athos.rutgers.edu (Charles Hedrick) writes: >It is quite true that a complete CL is beyond the ability of the >typical university hacker to produce. However if you start with Spice >Lisp, producing a CL is probably not much worse than any other Lisp >dialect. It's certainly within the realm of a team with a few good >people in it. It doesn't need anything like the resources that were >used to produce some recent big systems, e.g. X. However if you had >to do it from scratch (i.e. without Spice Lisp), and you had to do it >within a couple of years (Spice Lisp took far longer to produce >something complete - which is why so many CL's based on Spice Lisp >started out as half-baked), it would probably require a team that only >a couple of big-name institutions could put together. It's a fact of >life that software is getting bigger. It's hard for one guy to >produce a major package these days (except if he's RMS). While I generally agree with what you have said, (CL is not exactly easy to implement), I do not agree that CL requires two years with a team from big-name institutions. KCL was implemented by four people in about 6 months. Two were from kyoto (obviously), and two were from NDG (nippon-data general). They did not start from spice lisp. I believe that the time it takes for a project requires to a large part on decisions made early on in the project. The kcl team wanted to implement CL quickly and portably, and that is what they got. -- Oh yea, while Im at it, I believe that both KCL and Data General were left out of the master list of common lisp implementations. -- -- jeff seismo!godot.think.com!mincy
sfk@otter.hple.hp.com (Stephen Knight) (01/30/88)
We have been using the Poplog Common Lisp & Prolog systems for a couple of
years in HP Labs, Bristol. The Common Lisp is of a high standard, a bit slow
in places, but very safe to use. Prolog has speeded up a good deal since
Richard last used it [although I doubt it competes with Quintus
as a delivery environment.]
I have spent a fair amount of time in interfacing systems written in C, Lisp,
Prolog, and Pop11 together in Poplog. The ability to knit software written
in different languages together makes it a really neat system in practice.
Doing things like this is not always simple, but it is about as straightforward
as you could hope it to be when you have languages based on such widely
differing architectures.
For functional programming people, there's now a Standard ML system available,
although the hooks to interface it to the other languages are still being
worked on. I've used this a little and been quite intrigued by the
good performance and implicit polymorphic typing.
Here are the US contact addresses for Poplog for anyone who is interested
in following it up.
It's Robin Popplestone who handles the distribution for academic users, who
can be contacted at
Robin Popplestone
Department of Computer and Information Science
Lederle Graduate Reseach Center
University of Massachusetts
Amherst, MA 01003
email: pop@edu.umass.cs
And for commercial users, it is
Systems Designers International Inc
Industrial Systems Division
555, Read's Way
New Castle, DE 19720
USA
tel: (302) 323-1900
krulwich@kangaroo..arpa (Bruce Krulwich) (02/03/88)
In article <5084@well.UUCP> jjacobs@well.UUCP (Jeffrey Jacobs) writes: >Ok! Let's look at the facts; I hope that some of the rest of the net >will contribute information on whatever they are doing. Let's also >keep "Lisp work" confined to development and implementation of Common >Lisp, *not* applications, editors or what have you. Limitting your outlook to development of LISP systems is a rediculous way to judge Common LISP's acceptance in the CS world. The purpose of a language is to be used, in this case to be used largely for AI programming. When I was applying for jobs in AI a year ago almost every job I looked into was doing its development in Common LISP. As I recall, only 2 out of more than a dozen were doing it in anything else. These places were a crossection of industry research, applications, and pure research. If Common LISP has become the language of choice for AI research/programming, what is the difference how profittable its developers are?? Bruce Krulwich Net-mail: krulwich@{yale.arpa, cs.yale.edu, yalecs.bitnet, yale.UUCP} {harvard, decvax, cmcl2}!yale!krulwich Goal in life: to sit on a quiet beach solving math problems for a quarter and soaking in the rays. (ala MIP) Any B-CC'ers or JDS'ers out there??