cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nicola Ken Barozzi <nicola...@apache.org>
Subject [Proposal] Add pass-through capability to mounted pipelines
Date Mon, 30 Aug 2004 06:53:12 GMT

Summary:

Developing Forrest, I seem to have stumbled in the need of mounted
pipelines that do not halt processing if a match is not found.

Thus I propose that we add the pass through='true|[false]' attribute to 
make it possible for mounts not to necessarily halt processing if no 
match is found.

                             -=--=--=--=--=-
Rationale:

It's typical to use the URL space to partition *processing*; because of 
this it's natural to match a URL in the base sitemap and delegate 
subURLs to mounted sitemaps.

In Forrest we have decided not to do this, and to keep the whole URI 
space free for uses to fill with their semantical partitionings. To 
decide how to process a file we use file extensions and eventually
peek inside to take a look at the DTD.  This makes it impractical to
partition the sitemap in several mounted sitemaps.

It shifted from impractical to impossible as soon we tried to make it 
possible for users to add their sitemap to the processing. There is no 
clean way i know of ATM for me to delegate a part of the matches to an 
external sitemap without giving away a whole URL space. ATM we are 
asking users to *copy* the sitemap they want to use and change it, but 
it has already proved over time to be extremely fragile for updates. We 
do not want to keep such a lame contract with our users.

I hoped that calling the user pipeline with the cocoon: protocol and 
using a resource-exists could route round the issue, but it's still not 
ok. If a match is found, I have to process the pipeline two times, or 
rely on caching, which is not always possible or desirable.

So it seems that the only viable solution is to make it possible for 
mounted pipelines to not fail processing if a match is not found, and 
have the base sitemap resume processing. This is in line with what 
resources do now (in an unwanted but nevertheless welcome and happily 
used bug). James also has a similar thing in the mail pipeline matches, 
that can <passThrough>.

Thus:

I propose that we add the passthrough='true|[false]' attribute to make 
it possible for mounts not to necessarily halt processing if no match is 
found.

The proposed element for it is the <mount> node, so that the base 
sitemap has the say on whether it wants to make the subsitemap 
completely responsible or not of the subsequent processing. The default 
behavior would be identical to what we have now.

WDOT?

-- 
Nicola Ken Barozzi                   nicolaken@apache.org
             - verba volant, scripta manent -
    (discussions get forgotten, just code remains)
---------------------------------------------------------------------


Mime
View raw message