lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Erik Hatcher" <>
Subject Re: Contributor Document class repository proposal
Date Sat, 01 Dec 2001 03:17:36 GMT

----- Original Message -----
From: "Doug Cutting" <>
> I like the proposal, but I am not sure that Jakarta should host the
> contributions.  That starts to imply that they're a part of the Lucene
> project, and might be supported by it, etc.

Ant has a ton of "optional" tasks that it hosts and maintains.  Some are
genuinely useful to all, and others are a pain for the ant-dev team to
maintain because they themselves don't use them or the API's they are

> Is that okay?  Is there an important reason that you think these
> contributions should be hosted by Jakarta?

So certainly lucene-dev should be cautious about what the fold in, but by
including very common Document generators as part (as an optional download)
of the Lucene offering it makes it even easier for folks to come up to speed
with it and integrate it into their world.  Something that builds a Document
from a text file is a good candidate (although the demo FileDocument
suffices), and of course everyone and their brother wants to index HTML
files, and of course XML.  PDF is hot.  MS Word would be awesome (anyone
done this?).  Folks would salivate just to be able to download Lucene and
point it at a directory tree and have it indexed without having to write any
code - and only when they needed something custom would they then dig deeper
and learn its API.

Folks don't to know how to develop Ant tasks to make it work for them - just
as an analogy.  Struts - now thats somewhat of a different story!  :)

> Of course some contributions should be added directly to Lucene.  I'm
> hesitant to add contributions that require libraries that Lucene is not
> already dependent on, or that are not included in the standard JDK.

Yes, dependencies are tricky to deal with in this environment, but expected.
By utilizing <available> in Ant you could conditionally turn on/off the
compilation of code in the 'contrib' area without anyone being hassled - if
they want to build something specific, they have to know to get the
dependency stuff.  Lucene's nightly build with Gump could easily have
dependencies like JTidy added to it.

Thats the end of my sales pitch to get 'contrib' added to Lucene's CVS!  :)

I'm happy we've gotten the positive response that has been received so
far.... Lucene rocks.


To unsubscribe, e-mail:   <>
For additional commands, e-mail: <>

View raw message