-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi,
bit late to the party -- busy finishing a web-log analyses tool -- but here goes... Bertrand's tests ================ Reza pretty much covered every angle explaining the results Bertrand got with his test. I would like to add : The test data contains a *lot* of user-agent strings of devices which no longer occur in the OpenDdr resources. Using 'my' cumulative resources significantly. On the other hand for a large percentage of new devices in the OpenDdr data we do not have a user-agent string in the test data. Hence my plea for 'new' user-agents, to which we all saw the response... Release ======= Compared to all the 'others' DeviceMapClient is vastly superior : - - it's *much* faster, - - it doesn't require tons of code and hard-coded patterns so much easier to maintain, - - it has no dependencies so *much* easier to integrate...anywhere So, I feel confident releasing it and take 'care' of the .Net version(s) . But... Resources ========= "Aye, there's the rub..." I'm totally with Reza on the n-gram vs pattern (regex) thing. However, it will require : - - either OpenDdr 'maintains' n-gram resources. The problem with that, apart from finding them willing to do so, is that unless they maintain both n-gram and patterns, the 'old' OpenDdr code will no longer work - - or we maintain an n-gram version of the OpenDdr resources -- I don't mind investing time and resources in this : setting up the infa to deal with that and evolve to an automated system as proposed by Werner. Please all note that Werner updated the Resources to version 1.22 ! So, pending consensus : - - I don't mind taking care of the .Net release - - I don't mind taking care of the n-gram maintenance also : - - I'll update the test data with user-agent string I 'harvested' - - I think we need more 'new' user-agent strings and look into creating 'specific' test data sets, something I'm also up for -- I've got some experience in that area :-) So, with the resources updated, I think we're good to move ahead with an initial release and tackling the n-gram issue. Nice ! esjr -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJSaOtbAAoJEOxywXcFLKYcI5MH/iH/tg9fHZGYiVer82i2NYXS mgYK7Et62dUcchhmUHDtOQPf40wn/Td31AFyZRAxSt2xf2GwA77N/5hjIBOoVf1Y MVYhUEzG6KKOUIyfW16Rl8HOgJy4/GSQy5V6gTGag8TXoHxU+fGHW0KqU78d/D4C 72KjS3aXvXkRD+lVqA3SuQI8JR5URuQ43QMJY/mmdwEbfjg0dPCFjUxdOFXoo4KT YnCjNroHIOAmSlC0+FsqMBgLfnSiOPr2XZg+ksytVvC/Idf/79XC/6lW7gf3mDF1 /sk59DR5GNE8IE+VA6G8AZEp+5IN6uO8AjxooSNQ3/+lLntO7Urh5ZJzBmYwQiQ= =PyBb -----END PGP SIGNATURE-----
