wnewman is currently certified at Master level.

Name: William Newman
Member since: 2000-05-07 12:21:56
Last Login: 2006-12-15 13:31:11

FOAF RDF Share This


My formal qualifications are mostly in theoretical chemistry (QM). My work history is mostly in web and internet telephony. My most active programming interests are programming the game of Go, and developing and maintaining the SBCL programming system (http://sbcl.sourceforge.net).

Recent blog entries by wnewman

Syndication: RSS 2.0

I had a frustrating morning with Ubuntu Dapper and my new printer. And I get up early, and "morning" in this case means that "lp /etc/hosts" didn't work until after 11:45 AM.

First, "what he said" about bad interfaces, bad documentation, and how not to tweak an upstream software package. This was a significant part of how I wasted my morning. I encountered a web page like the one in his screenshot, I read the text "These web pages allow you to monitor your printers and jobs as well as perform system administration tasks," I hit the Administration tab, and it was all downhill from there.

Second, O'Reilly didn't cover themselves with glory either, unfortunately. I was using their Ubuntu Hacks book, and as I understand it in retrospect, much of my confusion was caused by following their instructions to rename the printer device from the autogenerated name like "/dev/usblp0" to something permanent like "/dev/usblp-epson-c88+" (so that one's printer configuration won't get mangled when you plug in a new USB device so numbers change). Sound idea in principle, perhaps, but apparently untested in practice (?!), because I find that CUPS can't handle it (and thus the following Ubuntu Hacks instructions end up failing mysteriously). And it's also implemented as a disturbingly flaky recipe ("chmod 777" on an "/etc/" file without explanation, and set "NAME{all_partitions}==..." even for printers which, AFAIK, don't have partitions). RELEVANCE TO ADVOGATO TRUST METRIC IDEAS ALERT: In a random net HOWTO that recipe flakiness would've caused me to search for another source of guidance; I could have saved myself a lot of grief had I applied the same skepticism to the book.:-(

Third, besides what he said ( pipitas' gentle chiding like "Dapper maintainers have crippled the CUPS 1.2 web interface"), my own tentative observation is that the Debian or gnome maintainers replacement for the CUPS 1.2 web interface that they crippled seems to be a distinctly less mature piece of software. In particular, once I did finally get to the CUPS 1.2 web interface, I found that CUPS gives explanatory text for what fields like "Name", "Description", and "Location" mean. The Debian/gnome GUI dialog provides the same fields, but only the single-word names with no documentation (no explanatory text alongside the entryboxes as in CUPS, and no Help or About button on the dialog boxes either). This, I can report from this morning's experience, is a very unhelpful UI design decision: I ended up telling the system that the printer's "Location" was "/dev/usblp-epson-c88+", i.e., the filesystem location that I had created as per the misinstructions published by O'Reilly. After all, by the time I had double-checked all the earlier dialog boxes that the GUI system had presented (no documentation there either, though admittedly they were remarkably shiny, with nice pictures of printers and all), there seemed to be no other place for the "/dev/usblp-epson-c88+" information to go. (As above, it seems that the intended route for that information was not to go but to come: it is to be autodetected by some means which works for "/dev/usblp[0-9]" but not for the "/dev/usblp-epson-c88+" name I had set up as per the O'Reilly instructions.) It's a ton of work to get internal documentation right, and it's only human not to have it all right --- but it might also be wise to count the cost of replacing an NIH system which seems to do a good job of it with a homegrown system which seems to do an only-human job of it.

Now, I haven't done that much to write keeping-simple-stuff-simple text, so I'm not sure I have any constructive suggestions for O'Reilly, other than that they might want to do something at least should happen that their revenues suffer from folks mysteriously becoming hypercautious about using their books. But for the programmers I'd like to suggest, besides the usual good idea of documenting your interface, also the good idea of not hiding underlying documentation. (And, as per pipitas yet again, not radically changing the functionality of the upstream software while leaving its documentation (and implicit documentation, like interface layout) describing functionality which no longer exists and which inexplicably breaks when the user attempts to use it.)

Finally and incidentally, I would also observe that for a decade or so I have joined in the chorus of mockery of MSWindows for the misdesign theme where an administrator must hike over to the machine being administered. Now it looks almost as though designers along the Debian/gnome/whatever chain of responsibility seem to think this is a feature, not a bug. That is, in what seems to be slavish imitation of MSWindows design, they have arranged things so that the only well-supported way of doing printer administration is from a GUI tool which can only be invoked from the gnome window manager running on that machine. (Or, possibly, the GUI tool could be invoked from the command line, and thus remoted through X? But I don't know how one would do that, because my usual trick for finding command line names for the gnome tools by looking at their Help/About text doesn't work: as mentioned above, the shiny printer admin tool seems to be unmarred by redundant Help/About buttons.)

another comment on another Reddit link...

I participated in the 2004 ICFP contest (in the Lightning Division only, as the 1-man "team" Busman Holiday Club), and enjoyed it, and did tolerably well, which might give me a reason to rationalize a fondness for programming contests:-) so take my opinions with a grain of salt.

While of course it is true that "fast programming isn't the same as good programming" as Vanier complains first, I doubt that that is all that big an issue. Playing Chess or Go under tournament conditions is not the same as really deeply understanding the game, but in practice there is a pretty strong correlation between the two.

Too my mind, a bigger issue is that small programs are not the same as big programs, and it's hard to address or even express big-program challenges in a short time. Much of what we do in programming follows from big-program difficulties which vanish in small programs. I think several things that Vanier complains about follow from this. For example, Vanier sounds quite justified in criticizing a programmer who refused to see the problems of coupling the UI with the program logic; his programming contest skills are unlikely to save him in large programming projects. Like Vanier, I find small problems to be unsatisfyingly far away from the ordinary challenges of programming, perhaps like having a contest in Chess pawn endgames instead of Chess. Still, small-problem skills are important, not so easy to learn, and not universally distributed even among graduates of good CS programs, so a contest based on them doesn't seem pointless. (Also... It seems a little odd that Vanier says he likes obfuscated C programs, for the programs are required to be truly tiny, and I haven't noticed programming-in-the-large skills on display in any that I looked at. And, for that matter, Vanier even says he likes the ICFP contest too. De gustibus et de programmibus nondisputandum est.:-)

Beyond the lack of programming-in-the-large, what I find most unsatisfying about programming contests is that they are naturally organized as surprise challenges. A surprise challenge format solves various practical problems, but it naturally ends up favoring contestants who happened to be particularly prepared for that challenge. It's a bit like having a "[generic] sports contest" where you don't reveal the rules until the day of the contest. If you think you are testing for pure athletic ability without people having drilled specialized skills, I think you are probably wrong. Your choice of ultimate-frisbee-like rules, badminton-like rules, rowing-like rules, or whatever will in practice tend to favor people who've done similar things in the past. For example, in the ICFP 2004 contest I likely got a significant benefit from having written several stripped-down assemblers when I worked as a programmer in high school; if instead I had taken the C++ job at Mentor instead of my FORTHalike/assembler real-time-control job, then in the ICFP contest I'd've had to make up more of the architecture of my software on the fly, and while it's not a terribly hard architectural problem, in the Lightning Division saving time is worth a lot. The winning Dunkosmiloolump entry also seemed to suggest a strong influence of the team having constructed similar solutions before. This isn't intended as a strong criticism of such contests or contestants --- I am still pretty pleased with myself that I did well, and I was still awed by the Dunkosmiloolump entry and thus the programmers behind it --- but it is a reason not to weight success in such a contest too heavily. It's probably important to be a reasonably good programmer-in-the-small to do well, but I don't think being the best such programmer in the contest is anywhere near a sufficient condition to win. Good luck coming in first in a contest like ICFP 2004 if your qualification is that you are stunningly good at programming --- in the field of numerical solutions to PDEs, or wireless telephony voice compression, or graphics rendering engines, or some other field where you have never happened to work on programs close to the ones required. Programming demigod (and DFW neighbor and fellow short person) John Carmack would likely stomp me harder in a problem soluble with preprocessed quadtrees than in a problem soluble with Metropolis Monte Carlo.

Finally, Vanier writes near his conclusion "Therefore, my advice to programmers who want to improve their skill: skip the programming competitions and start a free software/open source project of your own. If you do that, you have a chance of becoming a truly good programmer, not just a glorified code-grinder." Perhaps Vanier and I agree, then: I would not advise spending a lot of effort on programming contests either! My disagreements are with other points in his article, and with his original thesis that "programming competitions are (for the most part) a bad thing." Success in cooking contests might not be a particularly good measure of one's skills for the day to day challenges of feeding many people a variety of tasty food on a limited budget of time and money, but does that make cooking contests for the most part a bad thing?

gobry: Yes, I agree with various criticisms of Common Lisp, including your point about lack of features in standard libraries that are taken for granted in more modern languages. (Other Lispers, and ex-Lispers, too. Note, e.g., that Guy Steele was a towering figure in CL standardization before he did that Java thing.) And, if someone had written an article which I basically liked and which bounced that high on Reddit praising CL without mentioning the things that worry me, I might well have written a careful-about-skipping-over-those-gotchas message in response.

I can't think of any convenient links of me cautioning someone else that way about CL advocacy in particular, but if you really really search, I'm fairly sure that more than once in a web-searchable place I have said in effect "whoa, there are good things about C++ [and STL]" when people got too enthusiastic about CL.

I have also been guilty of spontaneously criticizing CL myself from time to time. For example, in the micro-user's-manual of the library I just released, on the hack that one uses in SBCL to set print properties which apply only at debug time, because ANSI provides no way to set them so that they apply only to an appropriate stream like *DEBUG-IO*,

#+sbcl (dolist (bind '((*print-circle* . t)   ; Thank you, ANSI, for 
                       (*print-length* . 64)  ;  specifying *PRINT-foo* 
                       (*print-level* . 64))) ;   as global variables!
         (push bind

Or for any CL users sufficiently hard-core to be messing around with the "genesis" phase of SBCL, try this, venting about^W^Wdocumenting the wretched hack that I wrote to work around the fact that we can't portably allocate a single vector of approximately the same size as the machine's address space:

;;; KLUDGE: This implementation seems portable enough for our
;;; purposes, since realistically every modern implementation is
;;; likely to support vectors of at least 2^16 elements. But if you're
;;; masochistic enough to read this far into the contortions imposed
;;; on us by ANSI and the Lisp community, for daring to use the
;;; abstraction of a large linearly addressable memory space, which is
;;; after all only directly supported by the underlying hardware of at
;;; least 99% of the general-purpose computers in use today, then you
;;; may be titillated to hear that in fact this code isn't really
;;; portable, because as of sbcl-0.7.4 we need somewhat more than
;;; 16Mbytes to represent a core, and ANSI only guarantees that
;;; ARRAY-DIMENSION-LIMIT is not less than 1024. -- WHN 2002-06-13

I have even been guilty of criticizing the CL code that I write or maintain myself from time to time. Check out a copy of SBCL and grep for "KLUDGE" and "FIXME"; even today, a fair proportion of those are me.

(So while you may suspect I was indulging in CL bigotry or even sneaky anti-Python-the-scripting-language advocacy, I think I was basically being in character as a cranky critic of language and implementation gotchas in general.)

In case anyone is interested in the somewhat obscure problem "how could I generalize the concept of interning and sharing unique DAGs so that I can allow the interned graphs to contain cycles?" or in any other formulation of the graph minimization problem described in Guido Tack's online notes, feel free to check out my latest attempt to solve the problem as a Common Lisp prototype program, with an unreviewed preprint describing the basic algorithm.

(The problem comes up in various fields. One example is detecting the equivalence of arbitrarily complex cyclic types in programming languages. Nonincremental minimization algorithms have been known for a long time, I'm trying to do it incrementally.)

I have some comments on the recent http://www.defmacro.org/ramblings/fp.html article on functional programming, and I decided I might as well post them here and send the author a pointer rather than just putting them in an email.

(You are approaching a skirmish in the language wars! Hit *BACK* now!)

I like functional programming, and I liked the http://www.defmacro.org/ramblings/fp.html article, but I think he probably should present more disadvantages up front.

One of the reasons that I prefer Common Lisp to Haskell (and why others prefer ML variants to Haskell) is that there are things which are clumsy or slow to express in purely functional form. Okasaki's _Purely Functional Data Structures_ is a marvellous book well worth reading for other good reasons, but one smaller benefit of reading it is that some such things come through there.

One annoying performance issue for me is hash tables; in an imperative language, it's straightforward to use the idiom of hash index into a modifiable collection to get O(1) lookup. Good luck doing this in FP! And, if you tell people that they should rewrite their BDD packages for reliability and clarity and shareability in Haskell instead of C++, and tolerate the extra O(log2 1E7) performance hit of doing all their cache lookups in nice purely functional search trees instead of ugly imperative hash tables, you will encounter some sales resistance --- though perhaps not as much, in the eventual runout, as if you conceal this performance issue from them and let them discover it for themselves.:-|

The issue of making small incremental modifications to large indexed or cross-linked data structures isn't just an algorithms performance issue, it can also make programs difficult to read and think about. I have done a lot of work on programs to play the game of Go, where on each move a player places a stone on one of the 361 points on a board. At various times I have written complete programs in C++, CL, and Haskell. In the imperative languages there is some programming hassle involved in making the changes undoable (so that you can try a variation, then backtrack to the starting point to try another), whereas you get undoability for free in Haskell; but in Haskell I found considerably more hassle in trying to express the small changes without doing a deep copy of potentially very large cross-linked data structures. It seems to me that this is a fundamental issue, not just a symptom of my naivete about functional programming. It might not be a big issue for an Erlang telephone switch, because I expect most of the state in such a switch is tied to an individual call, interacts weakly if at all with the state of other calls, and goes *poof* when the call ends. But if you tried to write, say, a MMORPG server in a purely functional language, I would expect that the ongoing small modifications to the complicated shared global state of the world would be a source of programming pain.

Also, purely functional languages seem to be unusable without laziness (to create cycles) and the purely functional languages people have not convinced me, as a casual user, that their handling of laziness is completely ready for prime time in large hairy systems. The difficulty of debugging lazy code is a minor issue; the difficulty of bounding the performance (especially heap usage) of complicated lazy code is worrisome, a potential showstopper. I would be very nervous about planning to develop a large Haskell system for something complicated and not similar to existing FP software (perhaps the MMORPG server example) without doing some serious investigation into that. I think it might be easy to end up with a server which failed from heap exhaustion when lazy-variable placeholders held pointers into futures which "common sense" shows could never happen, but which aren't manifestly unreachable and so therefore aren't garbage collected. Both my intuition and my superficial reading of the mailing lists suggests that such bugs are not hard to create, and can be hard to test for and hard to debug. The existence of various large FP systems successfully "used in anger" is reassuring, but only incompletely so: it's not hard for me to come up with a story why from the ability of version control systems and telephone switches and compilers to manage this problem it does not follow that it's manageable for all systems.

19 older entries...


wnewman certified others as follows:

Others have certified wnewman as follows:

  • wnewman certified wnewman as Journeyer
  • mwh certified wnewman as Journeyer
  • dan certified wnewman as Master
  • pvaneynd certified wnewman as Journeyer
  • cmm certified wnewman as Master
  • tbmoore certified wnewman as Master
  • rjain certified wnewman as Master
  • crhodes certified wnewman as Master
  • sulaiman certified wnewman as Master
  • mdanish certified wnewman as Master
  • bytesplit certified wnewman as Master
  • sye certified wnewman as Journeyer
  • scrottie certified wnewman as Journeyer
  • water certified wnewman as Master
  • ingvar certified wnewman as Master
  • bmastenbrook certified wnewman as Master
  • xach certified wnewman as Master
  • cyrus certified wnewman as Master
  • agroteckmoder certified wnewman as Master
  • pthierry certified wnewman as Master

[ Certification disabled because you're not logged in. ]

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!

Share this page