Re: Two Myths of XML

>>>>> "bijan" == Bijan Parsia <bparsia-vEoVa6Jh+/J4piUD7e9S/[email protected]> writes:

  >> in very particular social and historical contexts, most often
  >> ones in which radical social change isn't on the agenda.

  bijan> Does the latter bit matter? Arguably it's false in *some*
  bijan> sense: techno cusps *are* put forth when certain large-scale
  bijan> social change is on the agenda...just not the sort that
  bijan> eliminates or weakens elites. I.e., elites instiute certain
  bijan> sorts of techno change *in order* to tighten socio/economic
  bijan> control...often in "extreme" (aka "radical") ways.

Good. Thanks.

  >> I have recently focused on two specific areas of computer
  >> technology: the [2]Semantic Web and, one of its enabling,

  bijan> Bah. Calling XML "enabling" just buys into the rhetoric. It's
  bijan> a consitutive tech, but only because of social decisions.

Yeah, I really just meant "weakly constitutive" in the sense that it
got chosen, not that it's necessary. Sloppiness.

  >> When XML advocates, like those at the [4]W3C, say that XML is
  >> ``open'' they mean, approximately, that it isn't a proprietary
  >> data format. Which means you can access XML created by
  >> Corporation A's tools with Corporation B, C, or D's tools.

  bijan> This isn't quite right, is it? XML isn't a dataformat
  bijan> *itself*, is it? Maybe well-formed XML is. But if we add DTDs
  bijan> or other schemas, then "XML" is a data format description
  bijan> language. (You know this, of course.)

Yes, "data format" isn't technically very accurate; but I was using it
as a concession to rhetoric and economical prose. 

  bijan> The reason I raise it is that DTDs can be copyrighted. In
  bijan> fact, they *are* copyrighted (as the specific expression of
  bijan> an idea). Thus, automatically, whenever you make a DTD you
  bijan> make a proprietory *format*.

  bijan> So there's at least three senses of "open", "open" as
  bijan> documented, "open" as democratic, and "open" as "open to
  bijan> anyone to use/change/etc." And at least two objects which may
  bijan> be open, the tool and the product. XML, the tool, is
  bijan> reasonably open in at least two senses. But what it generates
  bijan> can be closed in every sense.

Agreed.

  bijan> "most often"...*please* :)

Thx.

  >> in the brains of technical managers and marketers -- that is,
  >> people who need to understand something about XML, but who aren't
  >> really technically-inclined, and so don't really understand much,
  >> if anything, about XML.

  bijan> Eh..Really? I think the w3c buys a lot of its own hype, and
  bijan> what of the moronic trio? Or MPH? *Lots* of "geeky" types
  bijan> fall for the hype..indeed, we may be even *more*
  bijan> vulnerable. Of course, there's vested interests at stake:
  bijan> We're cooler if the magic is real. Things we want happen if
  bijan> the magic is real. We get loads of money, fame, prestige if
  bijan> the magic is real.

Yes, that was a "lie of convenience". Fixed to be a lot tougher.

  bijan> [snip]
  >> Let's set aside the almost unimaginable scope of the project
  >> McDermott implies when she suggests that one could use XML to
  >> link legislative activity to its subsequent results in federal
  >> government databases and information systems.

  bijan> And the fact that if it could be done, it'd could be done
  bijan> with SQL, HTML, etc. Magic #2 :)

Yes, yes. 

  >> I want to focus attention on the idea that there could be, as
  >> Kotok reports McDermott to have suggested, ``a standard
  >> government-wide XML vocabulary''. I am stunned that I should have
  >> to remind the head of OMB Watch that the US federal government
  >> may be the most staggeringly complex human institution in the
  >> history of staggeringly complex human institutions. Its scope and
  >> breadth and coverage is massive.

  bijan> And not just in *vocabulary*. Managing the *data*, however
  bijan> expressed, is daunting. Managine *extra* data layered on top,
  bijan> more so. Achieving the *encoding* is scary.

  >> The very idea that there could be a single XML schema covering
  >> every institutional information requirement of the US federal
  >> government is a perfect example of the second magical thinking
  >> about XML, namely, that XML makes possible things otherwise
  >> impossible. It does not or, properly speaking, it cannot and
  >> never will.

  bijan> Hmm. I don't feel the punch here. A modest XPointer thingy
  bijan> might do *some* of the job. I'm confused about what level of
  bijan> detail he's suggesting. I'm also confused why you're focusing
  bijan> on the difficulty of the *schema*. 

Because it's an absurdly magical suggestion. Because that's how Kotok
reported it. And because magical thinking seems to often clump around
the schemas; hence, HumanML.

	 The difficulty is,
  bijan> presumably, "making govt more accountable", or more narrowly,
  bijan> "linking the bills to thier results as represented in gov
  bijan> databases". 

Yes, of course. But these are seen to be plainly difficult by almost
everyone, right? What makes someone like McDermott able to suggest,
seriously, that these may be overcome just *is* magical thinking about
XML and schemas.

	 It's hard to see how a schema helps all *that*
  bijan> much. 

Of course, that's a separate and valid point. Even *if* you could get
a schema of the right scope, how does that help? More magic.

	 But it's easy to see how all that can be subverted
  bijan> (e.g., there *is* no clear relationship between bills and
  bijan> *much* of the data in databases, lies and mistakes are shot
  bijan> throughout, plus we just don't *understand* the relationship
  bijan> between much of this; 

I agree.

	 XML can't fix that lack...*except*
  bijan> maybe by being a vehicle for a model that permits queries
  bijan> that produce results that we can successfully
  bijan> interpret....note how little has to *do* with XML.)

  bijan> So there's two bits: That such a Schema is easy or possible
  bijan> (and what about it *is* possible or useful, and how to
  bijan> decided) and that such a schema would do what it was claimed
  bijan> to do.

Agreed. I was just focusing on the first bit; the sheer possibility
for economy's sake; if it's not *possible*, it doesn't much matter if
it would be helfpul or not. I can (I think) imagine several things
that would be immensely useful that are impossible; faster-than-light
travel, say. (Maybe not physically impossible, but a not horrible
stipulation.)

  bijan> Maybe there's a third version of magical thinking, a varient
  bijan> of the second: Making an XML vocab *is* making a useful
  bijan> model/datastructure/information system. We all know that
  bijan> making an XML vocab is easy...just "dream up" some tags. But
  bijan> even *encoding* existing models in XML is non-trivial (look
  bijan> at RDF/XML). Just *serializing* simple datastructures is
  bijan> fraught. 

	 XML provides *no* formalism or procedure for
  bijan> generating useful models of complex phenomena. Or even simple
  bijan> phenomena. So, coming up with an XML schema puts the cart
  bijan> before the horse.

  bijan> You get at this in the next paragragh, but it's interesting
  bijan> to note that XML structures...even with XML Schema...are not,
  bijan> in themselves all that representationally powerful.

Yes -- or, better in the present case, they are vastly less
representationally powerful than they are taken (or than they are
being sold) to be.


  >> (To understand why this is magical thinking, try to write in
  >> plain English prose -- or in any language of your choosing -- a
  >> nontrivial, interesting vocabulary for describing the information
  >> encoding requirements of the US federal government. Or, to make
  >> things easier, how about doing that for just the executive
  >> branch?  Or, again easier, just the Justice Department. Take care
  >> that your head doesn't explode in the process.)

  bijan> Heck, then *using* it. Think about how much broken *XML*
  bijan> there is out there, much less HTML. Even if we could *design*
  bijan> it, imagine trying to *train* everyone to use it properly.

Indeed. *Even if* we could get such a schema, and even if it would
*help*, there'd still be the massively non-trivial application of it,
which would "just" consist of millions of *perfect* decisions every
day... Sheesh.

So maybe another variant? "XML makes machines intelligent" -- the
straight-up AI thing. It does seem in the background here. McDermott
surely doesn't expect *humans* to be making millions of perfect
decisions daily; "the machines will just know... because of the
schema".

Maybe not another variant but a very common instance of #2 -- XML
makes impossible things possible because it makes the machines
*intelligent*. (Which is stupid since there's no reason even to think
that perfectly realized sci-fi AI could do what McDermott asks.)

  >> In fact McDermott's only rival for being made the canonical
  >> instance of the myth that XML schemas

  bijan> Schemas or XML? I think it's XML that has the magic,
  bijan> though...that a 'vocab is all you need' when historically
  bijan> that's just been manifest false is the "adding xml makes
  bijan> impossible things trivial" trope.

  bijan> [snip]
  >> Which desperately bespeaks the ancient human dream, as old as the
  >> first stirrings of civilization, of the perfect language, now
  >> rebirthed of technophilic parents. I'd suggest a careful reading
  >> of Umberto Eco's The Search for the Perfect Language if I thought
  >> it would help.

  bijan> But you haven't gotten the *good* bit was that this was
  bijan> *impossible* before XML and *XML* makes it possible.

Yes, as I said, the thing with bashing HumanML is the embarassment of
critical riches! There's just *so* much lunacy from which to choose!!

But I've already made this longer, so why not some good good quotage?

  bijan> [snip]
  >> The other facet of taking technology to be independent of the
  >> social context within which it always already operates is to
  >> misjudge ts its alliance and use. In other words, the bad guys
  >> always have the newest, bestest, fastest, powerfulest stuff, and
  >> they always have more of it, and they seemingly always employ the
  >> people who created it. And, generally, any tool can be used to
  >> impede social change as well as to foster it.

  bijan> Indeed, I'd actually say that it's *easier* to maintain
  bijan> status quo via technotricks than to overthrow it.

Yes. Momentum and all that.

  bijan> Excellent article Kendallboo. I think there's a bit of
  bijan> confusion in the mid bits, and, really, there are Splendid
  bijan> with a capital s HumanML quotes:

Thanks for the comments, Bij. I'm posting a second draft now.

Thanks.

Kendall