LinuxExpo 1999

My quick takes on LinuxExpo 1999. Note that this is the first conference of this particular species (Linux, or free-software-related) I've attended, so I didn't really know what to expect.

 

• The scheduling information was rather confusing. Business track? Conference? Technical Track? Extreme Linux? C'mon, guys, most of us are hackers, not rocket scientists.

  (One poor fellow skipped Toon Moene's (<toon at-sign moene.indiv.nluug.nl>) presentation because he thought attending it required having signed up for the Extreme Linux track, due to rather confusing references in the schedules.)

 

• Other than that, it was incredibly enjoyable, very informative, thoroughly energizing, very inspiring, and so on!

  It certainly exceeded my expectations, and those of my wife (Barb) as well.

 

• Very few people wore ties.

  I was one of those few, most days, but that was because I happened to have a Christmas tie with skating penguins on it, so I wore that, and got plenty of positive comments on it.

  (The "official" symbol of Linux is a penguin.)

 

• Academia was not as highly represented as I expected. That is, more of the attendees seemed to be of the corporate, or developer, variety.

 

• They (sponsors such as SGI) did a pretty good job keeping munchies around the first-floor conference area, thank goodness!

 

• The various parties ("receptions") that we attended were pretty good food-wise, but too noisy for us. The two we skipped (Redhat's) were, based on reports, probably the noisiest of all.

  The good news is, this seemed to be more than acceptable for most of the attendees, who were young, driven, and not quite as nerdy-looking as I'd come to expect (based on my own personal appearance circa 1980).

 

• The qmail tutorial focused mostly on installation, replacing Sendmail, etc., but I enjoyed it because, in the months since I installed it myself, its "spirit" has sufficiently osmoted into my brain to have caused this tutorial to make more sense to me.

  But I think most of it was difficult for my wife to understand, though she seemed to appreciate the qmail philosophy as explained by the tutorialist, Russ Nelson, with asides and subsequent explanations from myself.

 

• "Managing Security Threats" by Alex Yuriev was really something.

  For me, it was a blast from the past, helping me recall the various ways I used to "hack" into machines just to prove I could do it. (Once I got a real job, shortly after turning 18, I no longer had the time or inclination to do that sort of thing.)

  For Barb, it was, I think, a rather sobering introduction to the clever sorts of schemes people use, sometimes for reasons most sane people cannot comprehend (and therefore cannot predict), just to make trouble for others using Internet cracking techniques.

  For what it's worth, I thought Alex purposely avoided delving too deep into the details of how to thwart attacks, to avoid making it too easy for crackers (or "script kiddies", a term I think means roughly the same thing as "cracker", but which I hadn't heard until it was used during this tutorial) to learn new ways to pierce the security of various systems.

  I actually probed this a bit by asking, or, rather, pointing out, the need to avoid making a certain type of security measure too easy to spot. Alex acted almost as if it was something he hadn't really thought about, very kindly amplified my point to the audience, but I think he just didn't want to go into that much detail, especially since there's an obvious countermeasure crackers can use, in many cases, to circumvent the particular protection I suggested, which he didn't trouble (as those trying to be "alpha geeks" might, to trump somebody else's input) to point out.

 

• "Applications: Replacing the Microsoft Exchange Server with Linux-A Success Story" by Henning P. Schmiedhausen was quite enjoyable.

  The sequence of events, information on a nifty small computer that runs Linux, and advice (such as, paraphrased, "don't offer Linux as a solution, offer a solution that happens to include Linux") were quite enjoyable.

  I think my wife would have appreciated the qmail tutorial more had she heard Henning's presentation beforehand.

 

• "Clusters: The APEmille super-computer: Linux in Theoretical High-Energy Physics" by Davide Rossetti (an Alessandro Lonardo was listed as well, but I think it was Davide who made the whole presentation) was fascinating, humorous, and revealing.

  Fascinating, because the design was, at the lower level of the basic computing engine, highly reminiscent of the mini-supercomputers which I both documented and for which I helped maintain a Fortran computer back at a company called Numerix in the mid-1980's. And the overall design was quite cool. It'd be wonderful to have one of these beasts in my basement, fully populated, although it is targeted pretty tightly to a few problem domains I don't particularly care about (currently), like Quantum Chromo Dynamics (QCD).

  Humorous, because they'd gone to all this trouble to reinvent so many wheels. Not only did they start with their own computing engine, which, in contrast to all the other clustering presentations (you couldn't swing a dead cat around LinuxExpo without hitting a presentation about a cluster), seemed like the manifestation of a severe case of NIH (Not Invented Here) syndrome, but they invented their own toolchain and their own "Fortran-like" language as well. (It's been over 15 years since I heard anyone use the term "Fortran-like" in a positive way. Fortran definitely has an important role to play, but hardly anyone seriously talks about designing new languages that look like Fortran unless they're backwards-compatible and, therefore, can run legacy codes without changes.)

  Revealing, because it seemed pretty clear, though not stated outright, that this project was funded not so much to provide a solution at an appropriate cost in an appropriate timeframe, though it might conceivably have done that better than using off-the-shelf parts (like Alphas), but to provide a way to show that Italians can design really cool computers too.

  My take on that is, yes, clearly, these particular Italians sure seemed to know what they were doing when it came to the design and implementation. While I might question the goal of designing their own system nearly from scratch, I don't worry about people reinventing the wheel as much as do some of my coworkers.

  After all, it was Italy's money, not mine, and the end result, even if it was a system not nearly as cost-competitive as something available today (which, again, I can't say), the end result is that, now, it's clear more people (and nations) know how to invent, or reinvent, those wheels.

 

• Bob Young's keynote, "The Open Source Revolution is Over and the Revolutionaries Have Won", was great, though his delivery was not, in and of itself, highly inspirational. Not that there's anything wrong with that, and, for all I know, he "toned it down" for this, a more technically-astute, audience, as compared to the usual sort (media, corporate, management).

 

• "Applications: The Gnumeric Spreadsheet: A Testbed for Component Programming" (I think I've corrected this title versus the program) by Miguel de Icaza was probably the most purely enjoyable presentation I attended.

  However, having not attended the "Introduction to GTK and GNOME Programming" tutorial the day before, most of it went quite over my head.

  It was the style of presentation, the huge number (and enthusiasm) of the attendees, and the zeitgeist that I found so enjoyable, I was sorry Barb had chosen to attend "Getting Linux Accepted into a Business Environment" instead, even though she surely got a lot more useful (to her) information out of that.

  Also, it was nice to see Miguel de Icaza finally starting to come out of his shell.

 

• "Supercomputing with Linux/Alpha" (again, my take on the proper title) by Greg Lindahl was pretty good.

  This was another one of those clusters presentations, in that the talk was about building a powerful "computer" out of a whole bunch of fairly ordinary PC-class, or workstation-class, computers, hooked together via a network, running software explicitly constructed to take advantage of this architecture.

  The executive summary: Lindahl expected to beat traditional, delivered-to-your-doorstep supercomputers on the price/performance bottom line, but was surprised at how many times, and how emphatically, his clusters beat the "big boys" on performance alone.

  Oh, of course, this presentation also made the Alpha look pretty great, whether it's the older EV5/EV56 (21164) technology, or the frighteningly fast EV6 (21264) chips.

 

• Jim Gettys' "The Two Edged Sword" was interesting from a historical perspective, at least, though I found his explanation as to why he doesn't like the GNU Public License (GPL) among the approved Open Source licenses a bit hard to understand. (He said something about there not being incentives to work on GPL'ed software. Rather strange, given that he was attending a Linux, rather than a BSD, conference, Linux being the only UNIX clone licensed under the GPL. Oh well. Probably best he didn't go into more detail and thus invite a flamefest in the auditorium.)

 

• "The Future of the X Window System" was somewhat interesting, though, again, a bit out of my league, given how little I've paid attention to X since attending my first tutorial on it back around 1986 or so. (I've used it, but am presently using virtual consoles under Linux. They're fast. I do look forward to getting the kind of real estate X will give me.)

  Apparently this meeting was something of an Event, as it represented a (semi-)public announcement that this new organization, X.org, had "taken over" X from "The Open Group", which apparently had annoyed many by claiming future versions of X would no longer be quite so free.

  So, even though I was more than a bit clueless, the meeting was enjoyable, because there was much rejoicing.

 

• "Programming: IBM Research Jikes Compiler" by David Shields fortified some of my recent feelings about how to redesign the g77 front end, among other things, so was very helpful.

  Jikes is a Java-to-Bytecodes compiler written in C++. Most of the presentation was about the progression of the project towards an open-source-development model, the things they learned during this process, etc.

  While that aspect of it struck me as "yup, been there, done that, welcome to the party, wish you'd just GPL'ed it though", I found the technical aspects of the presentation to be highly consonant with what others, who have managed successful bazaar-style projects, have been saying about how to architect a code base to truly invite participation. (Suffice to say, g77 presently is a counterexample of this sort of design, to a fairly significant extent. Now, I'm no longer reticent to preserve the old design concepts I learned, which don't scale well to open-source development, thanks to presentations like this one.)

 

• "Ports: Linux/IA64: Preparing for the Next Millenium", by David Mosberger, was rewarding and informative.

  Basically, they're trying to (and seemingly succeeding at) making sure core components of the GNU/Linux system are available for the IA64 (aka Merced, also covering McKinley) from "day one" of public release, while keeping details of the IA64 under non-disclosure.

  I see an interesting tension there, especially given the consensus among the cognescenti that Merced won't be much of a competitor to Alpha.

  Specifically, while keeping the information secret prevents competitors cloning Merced earlier than otherwise, it also keeps the price/performance offered by Merced at a higher (more expensive) level, declining more slowly over time, because the Open Source community (such as the gcc developers) doesn't get the information it needs to make the chip perform really well, even though a few members of that community (under NDA) do.

  That aside, I'm impressed by the devotion of HP (and, presumably, Intel) to making sure this new architecture doesn't get born closed, and to GNU/Linux in general.

  Also, my impression of the IA64 architecture itself is that, indeed, it is not CISC, RISC, or VLIW, but something sufficiently new to deserve the meta-architecture (or whatever the term is) name given it, namely, EPIC.

  Having pretty thoroughly hacked on, and thus learned, machines of all three types, and being something of a "budding" system architect for over 10 years now, I recognize many of the aspects of the IA64 they are willing to talk about as being not only distinct from EPIC's most obviously close cousin, VLIW, but things I (and others) at Numerix actually contemplated so many years ago, at least in "blue-sky" mode.

  And, as I offered during the presentation, one of the disadvantages of VLIW is that a given incarnation of an architecture targets a pretty narrow range of silicon, in terms of price/performance, whereas EPIC, specifically, IA64, seems to offer much more promise of scaling well (without requiring recompiling) from the low-power, small-size CPUs needed starting, say, around 2005 to fairly high-end processors.

  That's the trick I see RISC as having been able to achieve, somewhat, due to things like superscalar design.

  Not that there won't someday be a place again for VLIW, since I believe it's entirely feasible that, someday, substantial portions of the industry will depend 100% on portable source code such that they'll be happy to buy whatever chips represent this year's best technology, on a price/performance basis at the price and performance points particular customers happen to care about, without having to worry about backwards compatibility in the ISA. Simply recompile and get going. But realizing the practical benefits of that dream are a long way off, even if 100% reliance on high-level source code is closer at hand.

  Another particularly cool thing I heard during the presentation was how they're avoiding typical "skew" between the hardware being implemented and the simulator for the hardware being used by the software folks They said they have one formal specification for the processor, from which is automatically generated the hardware specification (presumably in VHDL or Verilog), the simulator, and the documentation.

  I've used a very primitive, simple form of that kind of approach in g77 (the intrinsics "data base" in intrin.def) for similar reasons. It is so much easier to find and fix problems that way! Well done, guys.

 

• Lunch with Richard Henderson, DJ Delorie, and Jeff Law, of Cygnus Support, plus Toon Moene, was a kick! Seemed to take well over an hour, while we all chatted away incessantly about all sorts of things.

 

• "Open Source on the Desktop" by Eric Sink was highly worthwhile. Barb found it particularly energizing. She later purchased a sort of "preview" CD-ROM of the (GPL'ed!!) software Eric's company, AbiSource, produces, due mostly to being impressed with Eric and his presentation. (Hopefully we'll fire up this puppy later, after I install RH 6.0 and get X, perhaps GNOME as well, up and running.)

  Oh, I mentioned during the meeting about how I long-ago hesitated to "accept" the Apple license for using Macintosh software since it forbad use relating to nuclear technologies. I later learned lots of people don't know about, or remember, this prohibition, though it's possible my brain's addled. I'll try and look it up, but feel free to send me a URL to an appropriate source of information, to help anyone who thought I was kidding (and somehow find their way to this web page).

  (My point had been that, if I was even mildly nervous about having someone else decide whether I was involved in nuclear-weapons work, it's probably best to assume there'd be way too much widespread nervousness about free-for-noncommercial-use licenses being used for substantial portions of the hypothetical Windows-replacing Linux desktop. Someone else had asked Eric if he thought such licenses would be acceptable, so I put in my two cents. I know, hard to believe, since I hardly ever express an opinion about anything.)

 

• The g77 BOF went much better than I expected, especially as regards attendance. (We quickly ran out of GNU Fortran notepads, and gave away the last remaining T-shirt the next day. The discrepancy was partly due to having given the notepads out early on, but the T-shirts only after a few people left, near the end of the meeting. Oh, except, Barb took some addresses of people who want T-shirts sent, so never mind.)

  For more on the BOF, see my g77 page.

 

• Toon's "Programming: Predicting the Future with Linux Numerical Weather Forecasting" was our final stop on the LinuxExpo express.

  This talk started with systems of equations that looked inscrutable to me in the Proceedings, but which made more sense as Toon explained them "live".

  Much of the talk was about the history of g77 vis-a-vis improvements in its ability to optimize various sorts of chunks of programming code typical of Fortran applications like Toon's.

  Some rather interesting questions were posed by the audience during this talk, which suggested to me that the audience for high-performance Fortran compilers is wide, though perhaps buried pretty deeply and not particularly aware of its other "members", so to speak. (After all, this was a Linux conference, though perhaps all those clusters talks brought out the number crunchers who normally hide from public view.)

  This presentation was basically the last time I was seen wearing my GNU Fortran T-shirt, which I'd worn to all LinuxExpo events. More than a few people, upon seeing my T-shirt (or my wife's), claimed to be enthusiasts of some sort or another for Fortran. Some asked when GNU Fortran 90 would be ready, others asked how the project was doing, and a few said something like "I thought I was the only person here who cared about Fortran!".

  That being said, of the few booths in the exhibit hall that seemed always empty of visitors, the Numerical Algorithms Group Inc. (NAG) ranked pretty high for apparent lack of interest, based on my occasional forays past it (which did include one visit, stopping for literature).

 

• More on the exhibit hall, which seemed just the right size, perhaps on the small side to me, but apparently significantly bigger than last year's LinuxExpo:

   

  • AbiSource boldly used a cardboard cutout of a cartoon ant as one of its promotional displays.

    That's right, folks...a bug!

    That takes either guts, or an endearing lack of marketing savvy.

   

  • Alpha Processor, Inc. was one of many booths offering a raffle of some sort, but since the prize was a 600MHz Alpha (didn't note which processor), I actually entered this one for myself. (The only other one I entered was to help my wife win a flat-screen monitor. We already know we didn't win that.)

   

  • BitMover gets my prize for "most smug booth boys". I kinda liked that, actually, and look forward to seeing just how well their BitKeeper product meets our OSS development needs going forward. (How's that for management-speak? See what happens when I spend an entire week full-time with my wife?)

   

  • Brown Computer Company, whose booth I didn't visit, gets my vote for strangest writeup in the booklet. For example, "Put us in your address book for help with Linux" is a nice thought, but would perhaps work better if they'd listed some address other than "North and South America, and Asia", which I infer from their listing those places as where their clients live.

    Of course, there's always online searches, but how many "Brown Computer Company" listings there are, I shudder to think. (Folks, remember to pick unique names, to help those search engines run smoothly!)

   

  • Cygnus became the meeting-place for some of us EGCS hackers.

    And, loathe as I am to admit it in public, I actually spent a fair amount of time talking with Stan Cox about their proprietary (shiver!) code-development-visualization-whatever tool. It's even conceivable I'd buy it for my system, but only after I figure out what is already available (and in what state) as GPL'ed software, and whether I should just write one myself.

    (Poor Stan also got to listen to my latest rant about how important Cygnus is to the free-software community, though he probably didn't find that too upsetting.)

   

  • DCG Computers, Inc. had a booth I managed to miss entirely, which is too bad, because I wanted to thank them for their past support of my g77 work by contributing (along with others) some equipment towards the Alpha workstation that I've had now for something like a year.

   

  • The Free Software Foundation had some literature, but I couldn't seem to find the top-level architectural docs on GNOME and such that I think Miguel suggested would be there. That's okay, though, I have too much to read as it stands, and there's this thing called the Web, I hear....

   

  • Fujitsu Software Corporation caught my eye because their handouts focused on their Fortran, as well as C and C++, development tools. I talked with them a bit about Fujitsu supercomputers, which I know only by reputation.

   

  • Linux Journal was giving away copies of its latest issue, and all I had time to do was grab one and read it later. I was pleasantly surprised, and might actually subscribe, though, again, I already have too much to read!

   

  • Patmos International Corporation was one of the few "we have some very serious iron sitting here" booths that I actually visited for any length of time. Their machines sure look cool (and scary) enough!

   

  • The Portland Group, Inc. was staffed by a friendly fellow who seemed to appreciate my thanking his organization (specifically, Larry Meadows) for long-ago sending me a list of specific Fortran features and extensions my then-new g77 compiler didn't support, but which his Fortran compiler did (based on running his test suite).

    Maybe to some people, that kind of act seems obnoxious, but it's actually very friendly to help a "competitor" by providing a list of problems with their product via private email. (I put "competitor" in quotes because it wouldn't surprise me to learn that the availability of g77 actually increases sales of products like TPG's Fortran compiler. It seems unlikely that g77 is good enough to take many sales away, whereas it's mere existence perhaps reassures customers that Fortran isn't dead yet.)

   

  • Proven Software, Inc. has an accounting package for business that, conceivably, Barb and I might adopt for our own use. But we have many other packages and possibilities to investigate first. (The ideal solution would be a GPL'ed package allowing multi-user, partitioned access to a multi-repository data base on a GNU/Linux platform. Well, that's my ideal solution for everything, actually....)

    One really cool thing about their product is that it runs under either X or curses! For a guy like me who happily spends months and months doing serious development without firing up X, that really means something. I'm not sure what, though.

   

  • Softpro Books sold us a number of books, mostly of the O'Reilly variety, including one "Web Design in a Nutshell" for each of us! (I plan to someday write a blurb on the importance of weasels in marketing.)

   

  • Sparco.Com got our notice because of its display of Palm Pilot machines and literature. Barb's interested in a Palm Pilot of some sort.

   

  • SuSE was visited, I think, by Barb, but not by me. But, they do at least some things right; years ago, they offered to ship me their GNU/Linux distribution gratis (all the way from Europe) because, like so many others, I'd contributed to software (g77)they were including on their distribution.

    I've been receiving their distributions ever since, and since I haven't ever actually made use of any of them, it's high time I persuaded them to stop. (If anyone wants to come to my house and pick up unused SuSE distributions, please do! I also have some used and unused Redhat and old Craftworks distributions with which I'd be willing to part. Better than throwing stuff like that away!)

   

  • Tangram Enterprise Solutions has a product called Asset Insight that struck Barb as being possibly useful for her organization. At least, she wondered what they were currently using to track PC's, workstations, etc.

 

• The hotel (Sheraton) in which we stayed was within a two-minute walk of the conference. It was a bit put-upon with staffing and other problems, but the staff were generally quite friendly.

  (We ended up being treated to two free meals. One due to dinner not being delivered within the promised 30 minutes, which we had to ask for, and was delivered in the least-friendly fashion we encountered the whole week. The other was graciously and swiftly granted by our lunchtime waitress, after she realized she'd forgotten to get us our drinks during our buffet lunch. Someone should go hire her for lots of money to do public-relations for a high-tech firm or something.)

  Nice touches about the Sheraton include that they don't charge for 1-800 calls from the rooms. (The worst is when the hotels claim they don't, then explain in the fine print that they do when they see the number is some long distance service, such as Sprint, other than the one they provide, and for which they charge an arm and a leg.)

For more on the goings-on at LinuxExpo, see Linux Weekly News. (Thanks to Toon Moene for telling me about this site and suggesting I send them a URL to this writeup of mine, and to LWN for adding it to their site!)

Copyright (C) 1999 James Craig Burley, Software Craftsperson
Last modified 2007-07-10.