At 02:20 PM 07/09/02 +0200, Guido Adam wrote:
>That's true. The information, which header fields to generate is quiet 
>hidden in the docs.

I'll try and make the docs a bit clearer there.


>I just tested it again.
>swishdescription is _not_ generated (i.o.w. empty), if IndexContents _and_ 
>DefaultContents are missing in the configfile.
>You have to use DefaultContents at least. You can leave out IndexContents.

That's a bug.  Thanks.

>And the metatags are not read, if you leave out IndexContents.

Both metanames and propertynames work for me without IndexContents.  You
have to using a parser that knows how to extract out the metanames.  The
default parser is HTML if you do not specify a parser, and that will parse
<meta> tags only (not fake html <tag> meta tags).  If you had a header
Document-Type: TXT then it won't parse the metanames.

[hum, I think the default parser should be HTML2 if available]

>My database records contain html pages.
>
>Looks like the "Document-Type:" field is not read correctly by the indexer, 
>if you use the "-S prog" switch. The indexer should use that field and not 
>the filetype it extracts from the URL.

Check again.  If you have in the -S prog program's output:

  Path-Name: foo.html
  Document-Type: HTML2

and in your swish config you say:

  IndexContents TXT .html

it will still use the header specified in the prog's headers (HTML2), not
the TXT parser.


-- 
Bill Moseley
mailto:moseley@hank.org