Re: [9fans] Publish and be damned.

[Lucio De Re <lucio@proxima.alt.za>]

This is a rambling rant, I should really not think of 9fans as
alt.folklore.computers, but I have no Usenet access to speak of,
anyway :-)

Also, I can only re-read it so many times before I consign it to the
list.  So please don't blame me if it annoys you.

On Sat, Apr 21, 2001 at 01:06:43PM +0200, Boyd Roberts wrote:
>
> no, i use revision control religiously (automated by make).  i prefer
> RCS for text files.  SCCS is a catastrophe.  never used CVS; do you
> think they find me contracts where i could?
>
No idea.  But CVS was just a bunch of RCS-automating shell scripts,
until someone had the (brilliant, in my opinion) idea of rewriting it
in C.  They carried the RCS baggage all the way, but bad habits die
hard.  Not to say that I don't like RCS, quite the contrary, but CVS
diverged in scope/intent from RCS and should have been re-engineered
at that point.  It's never too late, but now there's too much
unlearning to do.

But seeing that anything Open Source has a public CVS repository, I
should imagine the only reason you won't get a contract involving CVS
is because you insist in getting paid.

> the url should point at the current version.  the author uses whatever
> revision control with the master version of the source, which is the
> url does not point to.  new releases are shipped to the url.
>
Give CVS a bit of your time.  It could look a whole lot like RCS,
except you can point the CVS "shell" to a remote repository and it
discourages locking.  I personally like the
	cvs cvs-opts rcs-command rcs-opts
syntax  (where cvs is in effect a command dispatcher) in preference to
RCS's namespace pollution, but YMMV.  Then you get all the revision
freezes and other options that make life a misery (hard to track, no
easy way to figure out what it is you're after, but powerful when you
do) and one thing I know I miss: the ability to duplicate the
repository for caching purposes.  Geoff Collyer would no doubt prefer
to remote mount it, but there are permission problems with that; and I
may be missing something, too - for all I know it lurks in the
undocumented bits.

> > Different schools.  A friend and I once tackled a moderately simple
> > text processing task in AWK and C respectively.  We were all
> > experienced programmers, but not proficient in AWK or C, just
> > differently so, hence the choice.
>
> text processing in C?  are you mad?  C is dreadful when it comes
> to smashing strings about.  you wind up building some library
> so that you don't have to worry about it or you declare:
>
>     char buf[N];
>
> which is all very fine until n > N bytes get stuffed into buf.
>
I like C and feel even more at home in some assemblers.  I don't
undertake monumental tasks and panic when 64Gig RAM is mentioned (I
wonder if GCC will grow segmentation for super-servers, so we can go
back to small- and large-model programming on the Intel architectures
:-)

> i go for the former, if i have to use C.  ~15 years ago i'd
> have to do it in C, 'cos a 1 mip 11/780 wasn't that fast
> (well, our's were).
>
Hey, you could use Snobol (never understood it, myself).  I use Awk a
lot, TCL even more, but I rattle off small C idioms in my sleep and I
just don't notice the waste of time.  Give or take inserting items at
the beginning of linked lists, I seldom have to spend much time
debugging my small endeavours.  And I still do development for MS-DOS.

> this is a bit painful to use, but is doesn't break, until
> you run out of memory [it's old, this code]:
>
What does it do?  Do I really have to read that?

> > We finished at about the same time, with identical results, and we
> > were both surprised :-)
>
> yeah, but they were toys:
>
>     ``a moderately simple text processing task''
>
Sure, but how hard does anything have to be?

> after the reading the 150 or so pages of the 4 MIME RFC's [in near
> despair] i was faced with the choice of C or awk to 'parse' who
> was sending what to whom.  C would've been months of work, but
> i worked out a way to give them [the rat squad] what they wanted
> with awk.  so it was days of reading mime rfc's, with thoughts
> varying between:
>
>     - they did WHAT!?!
>     - i'll gonna hunt 'em down and shoot 'em
>     - kill me NOW
>
> and then a coupla hours with awk.
>
I tried to convert RFC-822 stuff to Novell's messaging service (early
nineties, when the Internet was becoming viable, but enterprise e-mail
seemed to be heading for a number of proprietary protocols) and back
again, never could get that to work.  But another difference between
us was that Unix did not exist in my environment until I sought out a
copy of Xenix (heavens, I'm not even sure how to spell it anymore) and
blew most of our budget on a 286AT clone.

My earliest encounter with Unix tools in a useable PC environment was
the MKS stuff.  Things were small and simple then.  My love of Plan 9
may well be nostalgia for those days.  But I wasn't confronted by
signals, ioctls and other ugliness until much later.

As for MIME, a better alternative will have to surface before the end
of this year, if we're not to lose the Internet to SPAM.  And that
applies to RFC-821 and RFC-822, too.  Why don't you distil your
dislike of MIME into a constructive proposal for a working
alternative?

Note that MIME handling tools are available, even down to the MS-DOS
platform, and have been stable now for years.  Without quite wanting
to sound sacriligeous, it is a poor platform that can't handle MIME
without blood, sweat and tears.

> > That's fine where conversation is the primary communication channel.
>
> nope.  you have the code.  read it.  that's what it's there for.
>
But code has no memory.  Previous versions do.  And when contributions
are allowed to be undisciplined, it is advisable to be able to compare
versions.  There's no guarantee that the latest version is better than
all predecessors, quite the contrary, in fact.

> at SRC (in Palo Alto) and PRL (in Paris) we used a package
> management system [The Siphon] and this was back in the
> early 1990s:
>
>     Francis J. Prusker and Edward P. Wobber. The Siphon: Managing
>     distant replicated repositories. In Proceedings of the IEEE
>     Workshop on the Management of Replicated Data. IEEE, April 1990,
>     Also appeared as PRL Research Report 7. (PostScript).
>
>     ftp://ftp.digital.com/pub/DEC/PRL/research-reports/PRL-RR-7.ps.Z
>
> distributed bug fixing.  now if engineering could have just listened.
> i could tell you a story about ULTRIX engineering, but then i'd have
> to shoot you or compaq would have me for breech of confidentiality.
>
Again, we all know of better ways, we all can pick on weaknesses, but
we leave to the Linux community to produce useable tools (we _do_ copy
their device drivers, don't we?) while we criticise them for not
following the True Faith (TM).  The crux of the matter is that RCS and
CVS are readily available (RCS in all *BSDs and CVS from CVS.ORG).
Where is a practical implementation of Siphon?  Why is no-one using
it?

> > You could have been well rewarded for that, ...
>
> i was: got treated like shit by one of the researchers for
> weeks on end and then small thank you when he blew away
> some file or other.  a PhD in comp sci but can't master rm(1)?
>
Isn't that what you asked for?  Forgive me sounding off, but you
sure sound like you couldn't handle thanks, you would have to bitch
about the magnitude, as you do above.  I have a rather hermitical
friend I could easy model your personality on.  Neither of you want
to be liked in a conventional fashion, mankind has to figure out
your good points the hard way.  Unfortunately, we all have enough
trouble managing our own lives not to consume effort on really
challenging personalities like yours, and at the end of the day we're
all losers as a result.  But there 4 billion people on this planet,
each potentially as deserving as you, even if many of them use
Windows.

> > the first crisis the DOS
> > user encounters when moving to Unix is the inability to recover
> > deleted files, and a facility as you provided would have been a
> > godsend to many like me who had to learn the discipline of
> >
> > alias rm="rm -i"
>
> these people should be all given loaded, hair trigger .357 magnum
> colt pythons shipped with windows.  it'd either teach them a lesson
> about confirmations or improve the gene pool.  do knives come with
> a -i option?
>
Quite the contrary, as Bill Gates has demonstrated.  Or do you
believe in "final solutions"?  It is weird to compare Bill as the
exact opposite of some unmentionable dictator, but he's got the public
very well sussed.  Pity about those in the "emerging nations" who
could emerge if technology weren't becoming ever more demanding
instead of more accessible.

> for our needs IIRC we needed about 4Gb for 30 days and then it'd
> cycle.  slowly i 'stole' RA-90's [1Gb] by shifting stuff around.
> now 4Gb is nothing.
>
It isn't nothing.  It isn't even cheap, when you have to back it up in
preparation for certain failure possibly even within its warranty.  I
have the Inferno floppy-image release that won't run on a PC with an
IDE adapter, two 3B1s that could teach a lot to a schoolful of eager
children with a single 40Meg, MFM drive.  But what can I teach them
when entry level in society today is a knowledge of a Win9x version of
Word?

> ever heard of running vi(1) on an 11/23 with 7th Ed and RL-02s?
>
No, by the time I encountered vi(1), 20Meg MFM drives and 1Meg of RAM
were available.  But the best word processor I remember (very vaguely)
using, was WordPro 4 on the commodore, 8K RAM?

What about supercalc in 64K?  Unix spreadsheets never even grew to
that sophistication, and that would have sufficed to slow M$'s growth.
I believe that the PC's launch was caused by Lotus 1-2-3, the killer
application of the time.  Others disagree.  But "lack of vision" would
be the AT&T trademark, not M$'s.  Hell, Win 3.1 was written by Pascal
programmers!  But it matched expectations a lot better than anything
else on the market at the time.  The point is, theoretical tools may
be very well, but people want to _use_ the toy they spend money on.
Linux's momentum came from being able to play graphic games early in
its life.  Its antichrist (as in anti-microsoft) image also helped,
naturally, but that's a religious issue.

> i think you better check that 'cos IIRC:
>
>     SCCS came out of USG: a dreadful piece of junk
>
That's an opinion.  Yes, I didn't like it either, and I embraced RCS
as soon as I discovered it, but something needed to pioneer the field,
and nothing better caught on in between, perhaps because of
"intellectual property" issues.  SCCS wasn't even bundled in SCO
Xenix, if memory serves, but then SCO were certainly at odds with the
universe.

>     RCS was 'BSD' based: raw, small and kinda neat but useless for
>                          DNS zone file version #'s.
>
Yep, its single biggest flaw.  And a bitch to change.  I tried to
change the $RCSfile:$ keyword to $Logfile:$ (to be able to exchange
files with PVCS which a client had bought) and never, ever found out
where to do it (and I'm normally pretty good at this type of thing).

> i have a selective memory:
>
>   173000g
>
MS-DOS explains it all: command.com and fc.exe are plain useless, you
had to write your own tools, without the benefits of research Unix to
guide you.  So I still do it as my favourite revenue earning
occupation.  But, like you, I'm fast becoming obsolete.  And I resent
it as much as you do.

> > Please write a book!
>
> ``that's not been un-thought of...
>
So, @%#$!* get on and do it.

> so do what the subject line says.  get a rapid prototype, see how it goes,
> fix/improve it, loop.
>
No, I don't see why I can't just use existing tools.  CVS (sourceforge
notwithstanding) is the conventional wisdom, all we need to add is a
bunch of URLs and a few ACME programs to translate them to CVS
requests.  And, as I mentioned to rsc, a librarian and publisher, as
recommended by Fred Brooks.  I would prefer a more competent CVS user
than I, but I'll do it while it needs me.

> if i get some peace in the next few days, i might do some real work.
> 10 years, 7 days, ~20 hours in this 'police action'.  i have a firefight
> to attend to first, so the rest is on hold.
>
Make peace with the world instead of doing things to spite mankind,
then focus your anger on what is obviously a very sharp sense of
humour.  We'll all enjoy your contributions, then.

If you need to develop something, by all means do it, but it would
take something along the lines of:

	$ export CVSROOT=:pserver:boyd@cvs.proxima.alt.za:/export/repository
	$ cvs login
	$ cvs import boyd/code "BOYD-V_0_9" "CODE-V_0_9"

(hm, don't try it yet :-)
to post your stuff on the repository from your OpenBSD notebook.  I
will assign you CVS write access if you find it useful.

++L