Re: XML Schema usage statistics (WAS: Draft minutes of 2009-05-12 TAG weekly)

A comment on this thread from a spec technology development perspective ...
it seems that the days of holy wars between schema languages or over. There
is a certain level of acceptance for both RELAX NG, XSD and Schematron, and
new technologies just take all of them into account, like XProc: It defines
processing steps for all of them, and the XProc schema itself is defined as
DTD, XML Schema and RELAX NG.  I think it is time to accept this status quo
and recognize this perspective: schema languages are one means (amongs many)
for building technologies and applications; if certain user communities
prefer one over the other, so let it be.

Felix

2009/5/22 Rick Jelliffe <rjelliffe@allette.com.au>

> Mukul Gandhi wrote:
>
>> I think this is good modularity. It's just that I have a difference of
>> opinion..
>>
>>
>
> Having sections in a spec as sign of modularity? It is like a bad joke.
> Under that criterion, what technologies are not modular?
> The more some part of a technology is severable and has been severed, the
> more it can be called a module.
> For example, XSD 1.1 assertions could be modular. They do not effect any
> other part of the spec. They could be in a different namespace. They could
> have an independent specification with a different editor who never needed
> to confer with the Structures editor except on the interface points.
>
> The structures spec could say "At this point you can put in extensions, and
> here is how extensions add their PSVI contributions." The Assertions spec
> could say "I am one of those extensions, and here are my PSVI
> contributions."  And an implementation could say "I understand that
> extension" and a user could say "I am not interested in the PSVI
> contributions from that extension, I don't need to have software that
> bothers with it." And it could be managed. I don't see any of that.
>
> There is *one* modular part of XSD, that is the datatypes. And look at the
> fruits of that modularity: the datatypes can be used by other specs (ISO
> RELAX NG, for example uses it.)  And, I certainly agree that there are many
> sections to the spec that could be pulled out readily: KEY/ID/unique being
> the most obvious. But modularity is not primarily editorial.
>
> Cheers
> Rick Jelliffe
>
>