From STATUS... Filter I/O

[rb...@covalent.net]

On 20 Jul 2000 rbb@locus.apache.org wrote:

>            MsgID: <Pi...@koj.rkbloom.net>
>    	Status: Dirk +1, Ryan +1, Greg -1

What will it take to remove this veto?  The technical reasons given for
it, that I can find, were:

1)  The API is big and gross.  ---  This has been fixed.  The API is now
small and lean.

2)  This avoids BUFF and goes straight to the network, avoiding chunking
and character set translation.  --- This was part of the goal of
filtering, to remove the BUFF code.  I have already submitted the
beginnings of a chunking filter, and I expect char set translation to be
moved out into a module.

3)  The code uses malloc instead of palloc.  ---  We have had this
discussion now, and I think most of us agree malloc is the right choice
for the data.  It is possible that we could use palloc for the buckets and
the bucket_brigades (In fact, it is already used for the
bucket_brigades).  This can be changed at any time, and is not written in
stone by any means.

4)  All modules must be re-written to take advantage of filters.  ---
This is no longer true.  Most modules should be re-written to take full
advantage of filters, but any current module can use filters without
modification.

I am VERY interested in getting past this blockade, and committing some
filtering code either late this week, or early next week.  There are
currently two different patches in CVS tree about filtering.  They can
both be found at src/lib/apr/buckets.  Greg's patch definately has more
comments.  I will try to fix this situation today.

I truly believe that if we can just commit SOMETHING, we can make a lot of
progress very quickly with filtering.

Ryan

_______________________________________________________________________________
Ryan Bloom                        	rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------