Hi Reza, On Sat, Feb 16, 2013 at 5:14 PM, Reza <[email protected]> wrote:
> Would it be a good idea to possibly offer 2 versions of the DDR? One > version would be a "lite" version which would contain a subset of currently > relevant devices. The "full" version would contain the set of all devices > ever document and it could even contain a larger set of fields. This could > be done with a unified DDR by adding a special attribute which would flag > from one set to another (or just maintain another set of DDR files). > Certainly a good idea, as this would give users the option to use either version they prefer. > My concern is that we have dropped a lot of device knowledge. I understand > that this was done for accuracy reasons but (from the dclass perspective) > accuracy is not a function of data or pattern count size. One of the things > which made the dclass project powerful is that it could quickly and > accurately classify any device ever documented by the DDR, pretty much any > device ever made, in constant time. This is a very powerful attribute when > thought about beyond this project and this was one of the reasons I made it > decoupled from any specific data source. It would be nice to keep this true > moving forward. Adding the lite vs full designation would allow us to keep > the flexibility to identify the devices which are relevant in the current > ecosystem or implementation. > In case this goes through, we could front the new version as the standard one, but let people know that there is a larger, although incomplete and comparatively inconsistent, version available that they can try and use. > It also may we worth considering porting dclass from c to Java. I actually > do have an older Java version. This would give the project some of the > advance pattern matching algorithms in dclass and remove certain limitation > which likely exist in the current OpenDDR APIs. > The valid point here being that speed in device recognition is a very important issue for some users, so having an engine that provides a high speed API recognition is certainly something that will appeal to many (if not all) users. > Just my $0.02. > > Thanks, Reza > --- > Sent from Blackberry Bold 9900 Best, -- Rubén Romero
