Re: URI Pattern Syntax

>I have found it useful to have a pattern, call it %?, that matched '/*'
>or the empty string, as in

Is this in addition to %* or instead?

I'm willing to have any number of wildcard matching facilities added. I almost 
added %? to stand for a single characte match. I think we need to apply extreeme 
caution however and only require the minimum feature set we _need_.

I like Dave's suggestion because URLs are hierarchical and there is a big 
differnece between {/pub, /pub/, /pub/WWW} and {/public}. Is %? the best match 
given that ? is often used as a single character match?

Note that %* would still be needed for suffix rules since %*.html would be a 
very usefull pattern to match on in assertion exchange. "All the documents with 
an http suffix from server foo are content-type text/html".

How about we consider some of the other possibilities. I would like to grab a 
chunk of escaped % space for template matching use:-

%( %) %| %+	Enough to allow us to build regular expressions as level 3.

I don't think we should go ahead and define a template matcher that goes this 
far _yet_ however. I think we should hold the posibility open when the URI spec 
is revised. 

[I'm aware thant %* is usually kleen closure so maybee I'm arguing against
my case here. But * is more often used as a wildcard match than in full blown RE 
matching. I'm not sure that users are able to deal with REs outside people who 
edit all their files with sed.

If I were designing URIs from scratch I would reserve * as a character and force 
it to be escaped so that the user would type foo*.html for wildcard matches. But 
I'm not and we didn't and so it will look clunky :-(

I could well imagine a future version of a http server accepting wildcarded urls 
to facilitate directory searches. Maybee they will not be entirely hidden from 
users in MGET, assertion exchange and the like.


Received on Friday, 23 February 1996 13:11:46 UTC