"William A. Rowe, Jr." <[EMAIL PROTECTED]> writes:

[...]

>>+1.  Changing apreq to share a common file bucket wouldn't be a 
>>simple thing, 
>
> Common file, multiple tempfile buckets (each with an offset/len
> into that tempfile).  You couldn't continue to share a 'bucket',
> each bucket should be one discrete data stream.

Yup, that's what I meant to say :-)

>>because certain functions in the perl api expose the 
>>spool file's name, and people are expecting that to be per-upload
>>and hard-linkable.  
>
> That's very, very nonportable.  Put your 'download' on a different
> volume than the tempfile, and bam - broken.

We already trap error and do a file copy when perl's
internal link() fails.  However what users do with
the spooled filename is beyond our control.

> The mime 'contents' bucket would still potentially be encoded text.

The default behavior of apreq's mfd parser is to simply split
out the mime parts into separate data structures.  Users
can register hooks within apreq if they want the parser 
to process the contents a bit more thoroughly.

-- 
Joe Schaefer

Reply via email to