Hello again, Jarkko!
Jarkko Hietaniemi wrote:
> I don't much like the idea of encoding the names of communication
> endpoints to class names. Net::FTP::GatekeeperDecCom, HTTP::SlashdotOrg,
> what next? The knowledge of how to contact/parse the "session" of
> a service is *data* (however complex, it's still data), not code.
> And a class name is (usually) much closer to data, in my mind.
Well, I can't say I'm entirely comfortable with it myself, and I'm open
to suggestions.
To me, the three issues are this:
1) Each site represents it's data in a particular way -- yes, this is
still just data, as you point out, but there needs to be a practical way
to represent the "bag of tricks" necessary for that particular data
source in a consistent, expandable way (by other people besides just
me).
2) The "lineup" of the data sources (sites) that have been implemented
should be arbitrary and configurable on a per-application basis. By
"lineup" I refer to the failover redundancy list of these specific
classes that each class supports. If the first class fails, it trys the
next data source, and so on down the line.
3) It would be nice if user-contributed site-specific expansions were
available to the rest of us, even without my having to include it in the
main bundle.
If I were distributing an application, rather than modules, my first
instinct would be to distribute some sort of configuration data file
that described the characteristics of each site, although this does not
adequately address point #3. At least in this way, however, users could
add sites without having to add classes.
I suppose I took my queue from Finance::YahooQuote, which provides stock
quotes specifically from Yahoo.
Does anyone have any thoughts on how to best encapsulate the three
points I mention above without using source-specific classes? I'm all
for it. I might add, the primary function of the site-specific classes
is to override the urls() method of the generic class -- so whatever
solutions are proposed should be able to adequately (and safely) produce
code for that method.
Thanks,
Matt Sisk
[EMAIL PROTECTED]
