How did TimBL's speech at Python 10 go yesterday?

[Michael Chermside]

>> I agree, and apparently so does TimBL. He was careful to point out that 
>> you don't actually go out to the web each time to get module, rather 
>> that the actual code is "cached" locally and there needs to be something 
> 
> Note that there's a distinction between using a URI as a unique
> identifier and actually downloading something from it.  In the XML
> Namespaces spec, URIs are used in things like <element
> xmlns:rng="http://xml.mems-exchange.org/schema/">.  That URI can't be
> dereferenced (the hostname doesn't even exist) but it's never
> retrieved, so that doesn't matter.  I think TBL's suggestion was more
> along the lines of having unique identifiers for modules and versions,
> and not primarily importing via URL.
> 
> --amk                                                  (www.amk.ca)
> 


Well, sort of. Except that he ALSO pointed out that you probably DID 
want to have something (like the module) available at the URL. In fact, 
he surprised me by suggesting that it was probably good practice (NOT 
mandatory) to put a DTD or schema or something at the URL used to define 
XML namespaces -- surprising because I heard it from TimBL, not for the 
idea itself, which was always downright obvious.

He certainly talked about downloading the module to local storage and 
keeping it there as long as the user wants... clearly this implies that 
there's something out there to download.

-- Michael Chermside

PS: I intentionally used URL in my comments, because if the URI is a URN 

instead of a URL then it can't possibly be "downloaded" as there's 
noplace to connect to. Just trying to keep the terminology straight.