Thread: [Solved in JD2] Better filehoster's mirrors support
View Single Post
  #4  
Old 30.10.2011, 14:22
Arioch1
Guest
 
Posts: n/a
Default

Quote:
Originally Posted by remi View Post
1) The automatic file renaming methods of hosts aren't known as the hosts don't publish them.
True

Quote:
You need to know all the details before anyone can implement an antidote.
False.

Microsoft Office formats were not published, yet people managed to open 90% of documents.
You need some idea about name mangling, then you can made some implementation.
And later you can always learn more and do better.
I think many Hosters plugins are not based on published APIs but rather on HTML page parsing - and while they break sometimes, they work for most time. Though they do not possess complete API specification

Quote:
It's possible your final file name might read like ___________-________.
True.
But not always, yes ?
You can do Bayesian probability here, with underscores here just not counting neither pro nor against the match, only accounting for invariant letters.

There is also metadata, like file length, file timestamp, md5 - which may be or not be available for matcher algorithm, which always be based on probabilities rather than certain knowledge.

Quote:
That's why I've always insisted that customers should indicate the mirror files manually.
They should be able to override FP's and FN's, but that does not mean software is prohibited from making more or less informed guesses :-)

Quote:
4) jD doesn't support resuming downloads from another link/host.
and that is where it can be improved :-)
Reply With Quote