Older blog entries for dan (starting at number 26)

OK, the first cut of my CORBA for Impatient People document is out. Feedback solicited (dan at telent.net), but don't expect me to read any of it before I've been to bed.

[edited to fix broken URL]

I spent most of today trying to work out what it is about the recent "collaboration" debate that annoys me so much. I still haven't entirely pinpointed it, even after posting a followup to lilo's recent article, so I'm still irritated.

Maybe I should post an entire article about how modest I am.

Taking a slightly different tack on the whole CORBA/CL thing now: I stripped down the libIDL/C stuff so that all it does is print out sexps. Then i read said sexps in CL and create IR objects there

As it happens I'm not actually using the ORBit IR any more either. I'll switch back to it and see what breaks when I'm happier that my own code is working.

UK folks are hereby notified of and encouraged to check out the UKUUG Linux 2000 UK Linux Developers' Conference in July. This is nothing to do with the "other" Linux 2000 - except, I guess, for the year it happens in .... Provisional speakers here

:sockets Version 0.3 is out and works with SBCL (Steel Bank Common Lisp, forked from CMUCL about a year ago by one brave William Newman, and has a reproducable build process. This is good)

According to the CL standard, READ-SEQUENCE is supposed to read the number of bytes you ask for, not just block until it gets some bytes and read those. It seems that the only way to get traditional Unix-style "block until something is available and return it even if it's short" is to write it yourself using READ-CHAR followed by repeated READ-CHAR-NO-HANG. If this is true, CMUCL gets it wrong. Rewrote :sockets tests until they work even with a READ-SEQUENCE that gets it right.

I was right about having broken it for CMUCL, incidentally, but it was a trivial fix

dhd is getting sexpual urges ...

Ankh may be claiming that rms hurts the open source community, or may be claiming that some people might want to claim that. I should probably not rise to that, actually. But, for the record, I don't. I did laugh at (but didn't participate in) the esr dimwit-fest, on the other hand.

yakk, the weird diary ordering you observe is probably a result of people editing their past diary entries.

I read the non-coding contributions article and was going to follow up saying ``surely this is stunningly obvious?''. Am moderately taken aback to hear that in many projects it's not - but then, other people's projects probably have several orders of magnitude more users than mine. Yes. :sockets version 0.3 - download it and be the second user! I'll reply to your mail, promise ...

More socket stuff, of all things. We now have it working on SBCL (though I fear I've broken it for CMUCL in the process, but I can fix that after a break/food/sleep or whatever seems appropriate)

It has a neat Makefile ...

New release when it works again, unless I get an answer from ORBit people first.

I wonder how many of the people who certified me as blue back when I first got on would certify me as purple if they were doing it now.

Nope. I don't certify people who are already well-connected. So I wouldn't certify you at all if I were doing it now. Ner.

Spent a large portion of day in bed listening to the radio. Alternated the rest between writing weird lispish-syntax stuff to generate java source files (and the generator that does this) and aimlessly browsing employment web sites. And an hour messing about with the orbit IR, but no real progress to report.

(mjs: nothing personal, but I remember when you had few enough certs to notice when some were from people you didn't recognise. And actually, I did mail you at the time, but it bounced,)

No recent diary entries. Well, I lost my laptop over the weekend and it's the only machine I've got that has my advogato cookie on it.

Watch me get slightly more diligent about backing stuff up ...

So on Friday afternoon I found that ORBit-in-CVS has much more of the IR than 0.5.0 did - and built it, which requires more frobbing with libraries than would superficially seem necessary (cries of `no! f**k off! I have got popt installed, I built it not five minutes ago' were heard around the office). orbit-irfeed now blows up in a different place to the one it did before, but after reading relevant ORBit bits I find that it's actually my fault this time. I can't create interfaces which dont inherit from anything, which I guess is fair enough.

Tip for drunk Londoners: always use a licenced hackney carriage (black cab). That way, when you leave your computer in one you know where to find it again afterwards.

Tip for Londoners considering getting drunk: don't. Bleurgh

From jennv's list of possible ``male responses to women who don't feel welcome here''

'yes there are problems but there will always be problems. live with it'

That position is just so completely antithetical to hacker mentality (as far as I understand it anyway) that I must boggle.

Boggle

• 'yes it crashes a lot but that's normal. live with it'

• 'no, you can't have the source code. I realise that's inconvenient, but you'll just have to accept it'

It's fine to say "ok it's broken, but I don't have time to fix it now" (implied subtext "convince me it's worse than I thought it was"), but where do people get off on saying "it's broken and you must accept that too?

There have been a total of 2 downloads of :sockets that weren't me. Neither have sent me mail, so I have no idea if anyone's using it. (My assumption is `not', and the amount that depresses me is `minimal')

So I got kind of wrapped up in CORBA stuff. My old CLORB hack area turned out to still have client support for CMUCL, so rather than port it to :sockets (for which I probably ought to think of a better name) I decided to start playing with ORBit and the GNOME CORBA stuff. All hail Matthieu and Dirk-Jan for writing the useful document.

CLORB doesn't read IDL: it works entirely using an Interface Repository. ORBit has an IR these days, but no obvious way to get interfaces into it. It blows up if I try to feed it from the orbacus irfeed. So, task (1): take one copy of test-ir.c from the ORBit source, one chapter from the CORBA/IIOP spec, and set of libIDL documentation (which is actually just about OK when read in close conjunction with IDL.h) and make an orbit-irfeed. Work continues apace

In other news, my new vanity domain name arrived. I'm not actually vain enough to tell you what it is though