[
Lists Home |
Date Index |
Thread Index
]
And the MID guys went from a set of genCoded objects,
to a pointy-bracketed form of C++, back to genCoded
objects, then back to a framework of C++-lke components.
And that points directly to the problem: a sharable
framework of component definitions. HTML didn't succeed
because of the basis in SGML; it succeeded because
of libWWW and the free Mosaic browser. Had ANY SGML
browser maker at that time given away their browser
and source code prior to those two products becoming
widely available, and had they settled on a protocol
and naming convention for goto hyperlinks implemented
over the Internet TCP/IP layers, it is conceivable
all the names on the pantheon walls would be different.
Instead, those names ended up cleaning up the stables
after the horses left the barn. "Git thar fustest
with the mostest" as the confederate general taught.
Well... I don't know anymore. It seems that every
time we try to design that tower to heaven, God
gets in the way. Given the numbers of people
who have to be convinced, it is like selling
the war: be simplistic and loud and doon't offer
too many details. Also threaten people you've
already ensured are unpopular.
But digging among the artifacts of ancient history,
one can learn a lot about building arches that last
a long time, about forms of government that perish
from the face of the earth given a more ruthless
opponent, that garbage is the best source of details
about cultures, and that computers and even batteries
were conceived and built before there were wall plugs.
So I am anxious to see what John Cowan comes up with.
All I can say is that every attempt I remember to solve
the semantic association problem of applying heterogeneous
markup systems either invents yet another gencoded
programming language, or codes a framework of compiled objects from a
set of virtual classes. If you go down that second path,
consider UML or something more appropriate for abstraction
than XML.
len
-----Original Message-----
From: Ralph Ferris [mailto:ralphf@valicert.com]
Ten years back, Architectural Forms were proposed as a possible
"reconciliation" mechanism for separate efforts then underway to develop
DTDs for UNIX documentation. Ultimately, one of these approaches (and the
organization behind it) went away and a single DTD was adopted: DocBook.
A few years latter, during the craze for the Web but before Netscape and IE,
the OLIAS SGML browser team at HaL Computer systems - where DocBook
originated - added an HTTP client to OLIAS and used AF-based mapping to
render imported HTML documents.
These events illustrate one of the key issues - and reasons for early
resistance - to adopting the Architectural Form mechanism as a design tool.
If there's an agreement on a vocabulary at the start, there's no reason to
be "more meta" by using AFs. For AF-based transformations to start to show
their utility, widely adopted vocabularies had to come into use with some
reason to transform between them. Note that, in this case, the reason wasn't
to arrive at equivalent "semantics" - not possible since HTML supports a
lower level of information transfer than DocBook - but rather rendition.
The OLIAS work in effect brough the Web to SGML. Unfortunately, the
under-funded OLIAS work never achieved the importance it could and should
have, leaving SGML theorists to contend with, among other things, the
once-infamous Netscape BLINK tag. BLINK tags were hard to deal with in terms
of the "HTML is an application of SGML so the Web is really SGML" line SGML
proponents liked to use.
Actually, for a number of years ISO had an ambitious attempt on its books to
take AFs beyond the realm of "more meta" and transforms. A number of
presentations on AFs referred to them as being equivalent, conceptually at
least, to virtual base classes in C++. As a rule, this analogy didn't mean
much to those approaching markup from a purely tech writing standpoint,
while it was met with skepticism or outright derision by OO theorists. The
ISO "Standard Multimedia Scripting Effort" (SMSL) effort meant to use AFs to
define object classes that included methods - "true" object-oriented design.
What could we have done with SMSL? For one thing, the BLINK tag could have
been defined in terms of SMSL classes. Then messages could have been sent to
object-instantiations to turn the things off. There would have been other
applications no doubt, but the resources weren't available to move forward
with the effort and eventually SMSL was dropped from the ISO work-list.
|
