wishlist re MGET, DDDS, various 'about:' solutions


I don't know if this thread is meant to be related to 
"best practices" or "data access", but if it is, I would
appreciate considering some practical thoughts:

Often, there is a difference between what *would be best*
and what *is usable* or *will be used*. people are already
using different solutions such as link tags, client or server 
headers, some local convention, someday-hopefully-valid-rdf
-in-xhtml, xmp, etc (see e.g. [1] for dif. approaches). I 
would really like a best practice recommendation for this 
"metadata discovery" problem, but I think that there will
probably be no single solution. _If_ you are trying to 
formulate a "best practice", it should be a solution (or
set of solutions) that can be implemented by as many web 
developers as possible, as easy as possible (network effect

There are different types of web developers out there:
- people deploying static web documents ("static")
- python/php/asp/... developers ("scripts")
- application server developers ("appserver")

furthermore, people are using different environments:
- hosted webspace ("ftp only")
- hosted webspace+scripting ("ftp+scripts")
- dedicated servers with root access ("rootserver")

So if this new SemWeb phase's objective is the wide
deployment of semantic web applications (including the
metadata discovery feature), then IMHO any combination
of developer type and environment should be taken into
consideration. There has been so much discussion on this
list concerning MGET, and Patrick has lots of convincing
arguments against the other approaches. but people still
seem to not like it. this should be considered. we need
a solution that people are willing to implement. I 
personally like MGET a lot, because client-side 
implementation is very easy for a "scripts" developer.
The problem with MGET is that you need to replace Apache
and need to have root access. DDDS needs nameserver access
if I got it right. So these solutions will probably exclude
the majority of web developers/environments.

For a "best practice" document, I'd expect something along:

- "if you want to relate your static html with rdf, do ..."
- "if you can read incoming http headers, you may add... and
   use ... "
- "if you are familiar with url rewriting, do ..."
- "if you are writing a scutter, try a HEAD first, if there is
   no metadata header, try an MGET, then ..."

Even if you are going for a single recommended way to 
offer/publish metadata, I would appreciate some guidance for
the already deployed/short-term approaches (e.g. "don't use
embedded rdf, if you want valid xhtml", "html comments are 
also xml comments, don't use them to embed metadata", etc.)

hope that was helpful..


Benjamin Nowack

Kruppstr. 82-100
45145 Essen, Germany

[1] http://lists.w3.org/Archives/Public/www-rdf-interest/2004Feb/0163.html

Received on Wednesday, 10 March 2004 07:51:20 UTC