[Bioperl-l] HTTP response size check in Bio::Tools::Run::RemoteBlast

[Scott Markel]

Sometime last week NCBI made a change to the HTTP response
for remote BLAST requests.  Based on when my regressions
started to fail, I think it was on the 14th.

The if( $size > 1000 ) check in retrieve_blast() now passes
when it shouldn't, meaning that intermediate pages are assumed
to be final results.  I'm now seeing response sizes of just
under 2000 for the intermediate pages.  A customer of mine is
getting about the same.

If this check is changed to 2000, then we're back in business.
We can't make the number too big or we'll start missing small
result sets.  A request for a single BLASTp hit gives me a
result size of about 3400.

Has anyone else seen this problem?  Is this a reasonable fix
to propose?  I'm a little concerned that whatever the number
is, it's very susceptible to changes at NCBI.

Scott

-- 
Scott Markel, Ph.D.
Principal Bioinformatics Architect  email:  smarkel at scitegic.com
SciTegic Inc.                       mobile: +1 858 205 3653
9665 Chesapeake Drive, Suite 401    voice:  +1 858 279 8800, ext. 253
San Diego, CA 92123                 fax:    +1 858 279 8804
USA                                 web:    http://www.scitegic.com