Re: plans for reference management

Christian Zagrodnick wrote:
[...]
 >
 > It would be great if we could share some code parts. CMFLinkChecker is
 > not *that* tightly bound to Plone that it wouldn't possible. If would me
 > nice to have the generic Zope part factored out an then having a Plone
 > or Silva version on top. But even if it doesn't happen, we at least
 > share the lms, right :)
 >

  Looking at the code of the CMFLinkChecker I found lot of things I would
like to have implemented in a reference checker anyway, like the catalog
storing the links.

  The retriever registration is implemented via adapters in my case
(in the upcoming next release ...) but that does not make a big difference
to me, so I should be able to migrate simply.
  Well, maybe we can use Five to factor apart and put things together; I would
have to stop writing mails and go figure it out, however ;)

Roman Joost wrote:

 > In the "link checking point of view":
 >
 >   Well, I think that some people know the CMFLinkChecker already. So, a
 >   likewise called product name for Silva would be a nice move. But, you
 >   made the initial development of the SilvaReferenceChecker. If you think,
 >   better to rely on your product, because the Silva world recognizes this
 >   as a product for link checking, we should use this name.

Roman: You are the only person which so far gave any feedback to me
about really using the current product, so if the term "SilvaReferenceChecker"
is already coined in the community I failed to notice ;-)

 >
 > If you already consider to enhance this product to a full reference
 > management suite, I think "SilvaReferenceManager" would be more
 > suitable.

Well, I would like to enhance the product to full reference management
in theory but the current progress seems to indicate its a bit far away ;-)

It seems the term "link checker" is more common than "reference checker"
so no problem for me to rename it for a fresh start. If we try to carry
over larger parts of the CMFLinkChecker its quite fair to name it after it,
i.e. SilvaLinkChecker.

  What I would like is to keep / migrate the "reference check" of the
current product, because:

   a) checking internal references is easier to implement
      than external link checking
   b) its more reliable
   c) especially for Silva the results are more interesting.
      Just consider the case of linking to a non-published document.
      Treating the link just as an ordinary link does not give any error,
      as the requst returns a 200 delivering the "sorry, not visible yet"
      message.
      Treat it as an internal reference its possible to figure out the link
      is broken.

hope I made my point here ;-)

  On the implementation side we maybe need another level of indirection
between the link/reference and the code checking it because it obviosly
doe not make sense to feed internal references to a link checker ;-)


  Anyway what I plan to do now is first to finish a last maintenance
release of the SilvaReferenceChecker already containing a few architectural
improvements like (homegrown) adapters for reference retrieveing,
and then start to discuss strategies to carry over stuff to a common
product, especially where to start and with what.

  I hope to get the release out next week and will then try to
get back to the discussion.


Cheers.
Clemens