Kahunapule Michael Johnson
kahunapule at mpj.cx
Wed Apr 25 20:00:18 MST 2007
Chris Little wrote:
> I'm not totally convinced we should do that. When I prepare modules from
> OSIS docs, I always perform validation in an external validator.
> (Personally I use Oxygen, but there are also XML Spy, MSV, topologi,
> Xerces, etc.)
> Do people feel that incorporating a real validator would make osis2mod
> easier to use?
Actually it could make it harder, by pointing out things that should be
fixed, instead of blaming the developers of The Sword Project for
glitches that happen later. :-) Seriously, it could save the step of
validating with an external validator, and I suppose that would be
simpler-- especially if this is done as part of a process you are
setting up for someone else.
> It could potentially cause the filesize to jump dramatically, so would
> that be acceptable?
Yes. It would probably take up less room than a separate validator,
anyway. Of course, if you wanted to run it on OLPC resource constraints,
then this might be an issue.
> If we incorporate osis2mod into either front-ends or installmgr so that
> users could import OSIS documents directly into Sword, would that
> support or detract from the case for embedding a full validator?
I would think that would support the cause for adding more validation,
so that downstream processes have to face fewer strange conditions.
Of course, if you can take invalid OSIS and surmise the intentions of
the author anyway, that would probably beat extra validation. All you
need is a good mind-reading module. :-)
More information about the sword-devel