Questions/Remarks about rbcollins_filters branch

<This message is targetted primarly to Robert and to people who know
about the filters he's implementing>
Hi all,

I am interested in the filters implementation (rb_collins_filters
branch) and after a few hacks around this new piece of code, I have
some (may be stupid) questions to make things clearer in my mind:
* What is the exact flow of data (from sockets to store). Are the
filter functions (for example dochunk) called once or many times during
an object reading. I mean what is the exact scenario ? :
   - read a chunk -> pass it to filter 1 -> pass the result to filters 2
-> ...... -> pass the result to the last filter -> append to store
object -> goto read chunk.
or
    - read all the chunks -> pass all the chunks through filters -> put
in the store

* In a filter function, what is the meaning of returning TE_CHUNK_A/B ?

Now, here is my (unique) suggestion (for the moment):
* Sometimes, it is better to have a filter triggered by some condition
(depending on headers value, for example, a filter to transform txt
files to html would be triggered only of content type is "text/plain").
So it'll be interesting to have this possibilty in the API.

Thank you... It is a good work Rob...

--Moez


Received on Fri Jan 26 2001 - 07:21:53 MST