On Thu, Aug 4, 2011 at 5:26 PM, Patrick Ohly <[email protected]> wrote: > On Do, 2011-08-04 at 16:29 +0200, Chris Kühl wrote: >> On Wed, Aug 3, 2011 at 5:58 PM, Patrick Ohly <[email protected]> wrote: >> > On Mi, 2011-08-03 at 17:02 +0200, Chris Kühl wrote: >> >> Hi, >> >> >> >> I've been working on retrieving the immutable vendor and product id >> >> provided by the Bluetooth Device ID profile (DIP). See my >> >> bluetooth-device-id branch[1] for what's been pushed so far. >> >> >> >> Being that we now have a means of getting reliable device information >> >> it seems to me that in the cases where this information was >> >> attainable, a matching score is no longer needed. If the fingerprint >> >> matches the product name from the look-up table exactly (and it will >> >> because we've created the lookup table and the template fingerprints) >> >> we suggest just that one template instead of several. If we didn't get >> >> this product name from the look-up table then everything should behave >> >> as normal because we can't trust the user-modifiable device name >> >> string. >> >> >> >> Does this sound reasonable or are there other reasons you'd want more >> >> than one template to choose from even when you're sure which device >> >> you're dealing with? >> > >> > What about a Sony Ericcson phone which unknown product ID? We have two >> > templates for vendor=Sony Ericsson. >> > >> >> In this case it would work just as before. If the device supports the >> Device ID profile but it's not in our look-up table of known devices >> then we don't use it. We need the look-up table to get the product >> string. > > I was working based on the assumption that our look-up table of devices > and the corresponding entries in the templates will be very small. >
Yes this is small currently, but should grow. And I plan on trying to crowd source this information once the Desktop Summit is over. Of course as you state below this doesn't mean we know which template this should be assigned to. > Where would we get a full list of product IDs for all Nokia phones ever > published, for example? And even if we could get such a list, do we add > all of those names to the Nokia template, without confirmation that it > really works? As you said, we shouldn't. > > So the more realistic outcome is that we have to make decisions based on > just the vendor name. > Right. This is available without support for the Device Id profile, by using a portion of the mac address. >> There are 2 things that have to happen for the matching score to be >> unnecessary. >> >> 1) The device needs to be in the look-up table. >> 2) The template needs to have a fingerprint that matches the product >> string from the look-up table exactly. If we or the user has added it >> here then this should be the template to use. >> >> If both are not fulfilled then we do as we've been doing. > > Okay. In that case the code needs to use the vendor name from the Device > ID profile instead of the user-assigned device name in the traditional > fuzzy search. > Ok. Some of the names in the vendor list obtained from the Bluetooth SIG would not match our fingerprints. For example, Sony Ericsson is "Ericsson Technology Licensing." SE also has another entry that reads "Sony Ericsson Mobile Communications." We either need to rename them to match in the list or in the config templates. I say we change it in the list. Cheers, Chris _______________________________________________ SyncEvolution mailing list [email protected] http://lists.syncevolution.org/listinfo/syncevolution
