From: Eugene Leitl (email@example.com)
Date: Fri Aug 11 2000 - 21:59:51 PDT
Mark Day writes:
> > The problem is obviously that we have to uncouple ourselves from the
> > market (which driving standard balkanization by attempting to lock
> > user base), while still maintaining noninfinitesimal user and
> > developer base.
> Sounds great. Drop me a line when you figure out how to do it. ;-)
It has to be noncommercial, and it has to be linked to a few, ideally
2-3 persons which have demonstrated past integrity is aspects to above
goals. The methuselah-in-spe standards have to be succinct, formally
defined as well as be supplied with a demo implementation in a
widespread, standard language (say, ANSI C) and come with a rich
torture test suite. It should be possible to test implementations with
a simple "make test" from a test suite digitally signed by standard's
authors, and no one else. Part of the suite should be a simple
standalone syntax check program. Standards should be fully backwards
and forwards compatible, syntax eliminating ambiguity (ideally at
syntactic atom level). When a major standard shift occurs, there
should be scripts available automating conversion of old standards to
new with minimal amount of dropped bits.
It's far from being complete, but it's a good start.
ASCII is a good standard. I don't think UNICODE is a good standard.
If it comes to document rendering languages, (La)TeX might be a
fledgling candidate. (A subset of) PostScript could be a better
candidate, would it be better standartized. In a pinch, a compressed
bitmap would do, provided it has enough colours for antialiasing.
> This doesn't actually line up with what I was seeing earlier in this thread.
> Previously, various complaints were of the form that standard X wasn't
> properly implemented in enough different places.
Well, mine was a new complaint ;)
If X was KISS enough, and would need to be certified (by an
automatical process) in order to be called X (I can claim I'm X all
day long, but if the digital signature won't verify users won't
believe me), and would come with a demo implementation already,
> But of course, that's because there aren't enough people who believe that X
> adds enough value that they go and demand X from their vendor. And there
If it doesn't comply, it's not X.
> aren't any vendors who believe that they can use X as a competitive
> differentiator. [Actually, for some values of X, there probably aren't many
> people outside FoRK who even know that X exists.]
Commercial players are not interesting in keeping X stable. This has
been ad nauseam demonstated by Microsoft et al.
> So I'd claim "the public" or "the market" does just fine at stopping
> standards churn at approximately the point where there's little value being
> added for the additional complexity. We might argue about whether it's
A standard is a standard. You can't be 80% compliant, that's about as
sensible as claiming you're 80% pregnant.
> stopping at the right point. Over time, one gets used to the fact that it
> almost always stops with something uglier/sleazier/cheaper/nastier than
> available alternatives, and I try not to spend too much time thinking about
> what that probably implies about people. But I don't think it makes sense to
People are not homogenous. I believe I share values with at least 5%
of the IT-savvy populace. Question is whether this is enough
user/developer base, and whether they can maintain sustainable
as they want.
> act as though the public doesn't get at least an approximation of what it
> collectively wants.
Sorry, but I'm not public. As a scientist interested in viable digital
publishing, I need something which has a readability scale approaching
dead tree, preferably acid-free dead tree or microfilm. You need
sustainability in a medium to allow accretion of prestige. Things like
http://xxx.lanl.gov are nice, but fall short of the mark. An open free
electronic publication medium with the renommee of, say, Nature takes
time to build, especially considering violent opposition and pressure
from the side of the establishment.
This archive was generated by hypermail 2b29 : Fri Aug 11 2000 - 23:06:54 PDT