On Wed, 2002-07-31 00:46:01, Ace wrote:
> No, there's no shebang at the start... it begins with funny special 
> characters like umlauts, vocals with acccents... in fact you can take a 
> look yourself as the file is publicly available, as it seems.
> 
> http://www.rrze.uni-erlangen.de/institut/software/produkt/spss/allbus/allbus96.por
> http://www.rrze.uni-erlangen.de/institut/software/produkt/spss/allbus/allbus96.sav
> 
> So maybe the webserver delivers the wrong content type because the magic
> of mime_magic isn't magically enough.  I could of course just exclude any
> .por and .sav or only include known endings; it's just bad if the users
> only need to put in a file that is somehow broken to make the indexer
> crash... users are evil ;-)

That web server is arguably misconfigured, as are many web servers. It
returns "text/plain" as the content-type for files it does not recognize.
It should, in my humble opinion, return a content-type of
"application/octet-stream". If you have control of the web server
configuration and it's Apache look at the httpd.conf file and change the
DefaultType setting.

The file in question is a SPSS data file. There is a fair amount of plain
text in such files and it is conceivably useful to index them. Certainly
swish-e should handle non-text files in a robust predictable fashion.