[SCL] Re: XML syntax for CL

[Murray Altheim]

Tanel Tammet wrote:
> Hi,
> 
> A few words about the SCL-in-XML subproject.
> 
> I'll give a bit of history to help sorting up possible confusion first.
> 
> - A few weeks ago I sent an early draft of SCL-in-XML proposal
>   to the SCL list, as I had promised to do. Discussed it later with some
>   theorem proving people, got some ideas for improvements. Waited fo
>   SCL list comments.
> - SCL list was quiet, no comments, no traffic.

Yes, I saw the post and reaction.

> - Yesterday I asked Pat to look into the proposal and comment.
>   Pat promised to do so.
> - Now it appears Murray has taken up the draft proposal, worked on
>   it and directly published the new proposal at
>   http://purl.org/xcl/1.0/  and
>   http://www.altheim.com/specs/xcl/1.0/
>   claiming the author to be Murray Altheim.

To do otherwise would be incorrect and rather presumptuous. I'm
sorry that your email wasn't mentioned in the bibliography, and
there's currently no Acknowledgements section at all. I'll fix that.
I'll also explain below more on this subject.

> - Pat noticed and commented on the proposal, being basically
>   unhappy with the XML itself.
> - Murray sent a long letter explaining his background, motivation,
>   OK-ness of XML etc.
> 
> 
> First, many thanks to Murray for taking up the draft and improving
> on it!

You're quite welcome. As I mentioned in my long-winded reply, the
motivation in posting XCL was to influence the design of the committee
product towards satisfying (perhaps not in version 1, but in version 2)
some of my design goals, predominantly towards implementing CL-on-the-Web,
not simply CL-in-XML. I believe they're compatible, just different
versions.

> Considering the contents of Murrays and Pat's suggestions and ideas,
> I have to think a bit before coming up with a new revision of
> my draft. I'll probably do it in a day and send comments/answers
> along with a slightly modified draft.
> 
> Second, in order to avoid further confusion I'd propose that we
> proceed as follows (cannot force anyone to do so, but it just
> seems reasonable):

I'm happy to make these modifications if the general feeling between
you all is that this is the direction you'd like to go. You also
have your own work. I did not post XCL with any attempt to hijack
the committee's efforts, nor to force myself into your group as an
editor, just to be clear. You might also note that I wrote that
entire spec in a 24 hour period, which might explain why my last
message was such a dingie.

> - Murray, would you please edit your current document on the web,
>   saying directly at the title / authorship header part that:
> 
>   - it is an early draft

It says that quite clearly in the Status section, and in the DTD.

>   - it is a part of (or based on) the SCL project

It is based mostly on the CL project, in actuality. And influenced
mostly by conversations I've had with the CG, XTM and GXL communities,
and with John Sowa and others in private email.  I could't at the
time assume it was part of the SCL project, nor do I currently know
the relationship between the SCL project and the CL project. Which
was why I made sure I used a different name, so there'd be no
confusion. Apparently that didn't work out so well...

>   - editors are Murray Altheim and Tanel Tammet
>   - no authors should be mentioned (or the whole collective authorship
>     of SCL people plus Muray should be given)
> 
>   otherwise it looks like you hijacked the SCL work (inc mine) without
>   giving any credit.

I believe that is unfair. The References section does include CL
documents, and as I mentioned I am happy to include a reference
to your email. But you should note that I've been working on this
since the CL (not SCL) info was made public. Actually, I've been
working on this in effect since I first made an attempt to do the
same with CGIF, when Peter Becker and I collaborated on "XCG" in
2001. I have XCL and CLML DTDs going back many months. I noted
there was already a CLML so I chose XCL for my public proposal.

But to avoid any incorrect appearances, I'll add an Acknowledgements
section to clear up any misunderstandings. And a note later today
stating clearly that it is NOT the work of the SCL project. It was
a solo publication, the production of a single author. Nothing comes
out of a vacuum, but until there is some understanding between all
of us that XCL is to become part of a community endeavour, it would
be incorrect for me to make changes to its authorship (I'm not
opposed to that, but not until such time as a decision has been made).

Now, if you all want me as an editor, or co-editor, that's fine. I
just don't want anyone to think I'm demanding that. I do agree that
it's better to have one or two editors and then credit the authorship
in the Acknowledgements (i.e., "Authors: see Acknowledgments"). Editors
are responsible for the publication.

> - Murray, I'd be most happy if you (at least initially) join the
>   SCL-in-XML  draft work in the SCL context: help improving the
>   SCL-in-XML draft, for the SCL report.

I'd be happy to. Thanks for the invitation.

>   Once we get somewhere there it would be OK to make a branch
>   if you so wish, call it XCL etc. But not right now. Hopefully
>   we get somewhere with the draft very soon (after all, the
>   minimalist common logic is a very simple matter. The nontrivial
>   things are extensions, which we should IMHO put to a separate
>   layer.)

I would like to make sure that the two approaches are in harmony,
that one is a follow-on from the other syntactically. For example,
the web version uses a set of tokens *or* PSIs. The non-Web version
would only use tokens, and likely not be extensible. Just FOL. And
no XML namespaces, no "arbitrary" XML, as either would encourage
people to create non-SCL/non-XCL documents. The Microsoft extend-
and-destroy thing we need to avoid with all diligence.

But there's no reason why the current XCL proposal couldn't be
easily scaled back to a version 1, with a clear roadmap on how
to get to version 2.

> - Regardless of the verbosity of XML we need to have SCL-XML as one
>   of the concrete syntaxes. We cannot remove the inherent verbosity
>   of XML there. There is no point in complaining that XML is
>   verbose: it is designed to be so. Our ownly means of countering
>   would be to AVOID writing SCL-in-XML concrete syntax altogether,
>   which probably would not be a good idea.

Honestly, the only verbosity is visual. In terms of components, it's
not any more verbose than the various math notations. Tags are simply
delimiters for components in order to make explicit what humans see
implicitly in say, algebraic notation. For example,

    p(x,1)

is six components plus positioning: 1. a predicate whose 2. name is
"p"; 3. a term whose 4. name is "x"; 5. another term whose 6. name
is "1". The parens are delimiters.

    <pred name="p"> <term>x</term> <term>1</term> </pred>

but now the computer knows what is what, unambiguously. There's
now still six components, a <pred> element, its 'name' attribute,
plus two <term> elements and two CDATA strings "x" and "1". The
delimiters include their mnemonics, that's all. In XML we add info
about XML and any encoding, wrap it in a document element to
provide a namespace and a single root.

   <?xml version="1.0" encoding="UTF-8"?>
   <xcl xmlns="http://purl.org/xcl/1.0/">
      <pred name="p"> <term>x</term> <term>1</term> </pred>
   </xcl>

and now it starts to look thick.

But I agree, there's little point in complaining about this. It's
what computers understand, syntax errors are quickly flagged (where
in algebraic notation you usually just get a different formula),
and it will buy us a lot more visibility than nested parens.

> - XML and RDF/OWL family are two separate families of languages.
>   I am beginning to think that we might need a separate concrete
>   syntax for SCL-in-RDF (different from SCL-in-XML!).

This is likely to be so. But it's much easier to read non-RDF, and
much easier to convert "simple" XML into RDF. And I'm absolutely
certain that someone in the RDF community will do it anyway.

>   It does not seem to be feasible to use ONE concrete XML syntax
>   for SCL-in-XML and SCL-in-RDF. We should not try to treat
>   the current SCL-in-XML syntax as meant for RDF. It is not.
>   It should not be meant for RDF.

Agreed.

>   We might simply need another syntax for putting SCL into RDF.

Yes. Or an XSLT stylesheet to do so, though I'm not entirely
certain you can create RDF from XSLT. I think Nikita Ogievetsky
has converted XTM into some form of RDF, so it's likely.

>   If we do it, the result would conceptually be a direct competition to
>   OWL, btw.

Which OWL? For those not interested in OWL, it would not be competition,
but simply a reasonable alternative. I'd never use OWL. Am I in the
minority? I don't know.

> - The BEST WAY to improve the SCL-in-XML is to come up with an early
>   draft for SCL itself. We still lack it! Who is supposed to be
>   responsible for this draft? Just a brief, ca three pages draft would
>   IMH be sufficient for significant progress.

I can take the XCL spec as it is, make the mods you recommend, and
post it as a version 1.0 draft. It will be missing all the features
you don't want. I look at the current XCL proposal as a target for
CL-on-the-web, but it would not be difficult to strip the extras and
provide both (two separate specs, or one spec with two levels).

>   Otherwise we will unnecessarily concentrate on some concrete
>   syntax issues.

I would rather they be considered as decisions on goals. This group
really needs agreement on a set of requirements before it can approach
a syntax, abstract or concrete. But you're correct that we should
avoid discussing syntax details. My proposal was meant as goals and
ideas on features, not so much a specific syntax.

> - Regarding the XML verbosity argument, the best way to counter
>   is to come up with another SCL concrete syntax, be it S-expression
>   based or "classical"-look-like syntax.

That would be a different project, IMO.

>   Again, who is supposed to be responsible for bringing up a small
>   draft (again, just a few pages, mistakes all forgiven) for
>   SCL-in-S-expressions and SCL-in-classical-syntax?
> 
>   Anybody feeling some courage in yourself, please just write one
>   (even examples and a laundry list of main items will suffice)
>   and send it to the SCL list :-)
> 
> Regards,
>          Tanel Tammet
> 
> PS: please observe the name is Tanel, not Tamel :-)

Yes, again, sorry about that.

Thanks for considering my proposal. I'll make the credit modifications
to the XCL spec and an Acknowledgements section, but not to the
author/editor section until there's been some resolution on the way
forward, that that's the way everyone wants to proceed. I don't
consider myself an editor on a community project until that community
has made such a determination.

Cheers,

Murray

......................................................................
Murray Altheim                    http://kmi.open.ac.uk/people/murray/
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK               .