Hi, Darren, please see my answer inline. Darren J Moffat : > Shi-Ying Irene Huang wrote: >> Ubuntu bundles tracker as its default desktop search tool, and >> Suse bundles beagle as its default desktop search tool. >> Meta Tracker needs to be integrated into Solaris also since >> there is not desktop search tool. > > So why are we choosing tracker and not beagle ? I wouldn't hold up > the case for the answer but I'd like to know since I'd heard of beagle > before but not tracker. > Because beagle is a mono based application, which is a .Net runtime environment on Linux platform. As Brian said that, sun has no plan to integrate mono now.
>> 4. Technical Description: >> 4.1. Details: >> >> Meta Tracker's key features includes: >> * Extracts file contents as plain text and indexes them. >> * Extracts Meta data from file also, for example, >> Artist/Title from one mp3 file >> * Responds in real time to file system changes, automatically >> keeping its meta data database up to date. > > How does it do that on Solaris ? Tracker depends on gamin API to monitor file system changes. Please refer to LSARC/2007/398 for more informations. > > * What is the performance impact of doing this ? Tracker daemon itself is a light weight daemon. When its indexer to filter file, it always runs in low priority as "nice -n19". And user can set preference for "throttle" from 0 to 100. if throttle is bigger than 0, tracker will call g_usleep(throttle) to sleep current thread after indexing each file. > > * What is the impact on disks spinning down and cpu's clocking down by > adding this ? > I'm asking this because I've had bad experiences with spotlight > on MacOS X keeping disks spinning sometimes. > Our QA has tested tracker for some builds, it does not cost much when indexing for the first time. And after it, trackerd is very light weight. > * Does this work if the users home directory is NFS mounted ? Yes, it can index files on NFS directory also. And there is one option which can set in preference ------ "allow to index mounted directory". > > * Does it cross filesystem boundaries below the users home dir ? Yes. And Users can choose which directory to indexing. > > * If the users home dir (or subdirs below it) are ZFS datasets that > have the ZFS snapdir visible will this index the snapshot dirs too ? > Will it be confused by that or provide confusing output to users ? > > * Where does it store the, potentially highly sensitive, information ? It will save the indexed data in sqlite db format under home directory, the file permission for DB is 600. By default, tracker will ignore all directory begin with ".", for example, ".gnome". by this, it avoids indexing some sensitive data. And ZFS snapdir begins with ".", so it will be ignored. Pidgin is one exception. Pidgin stores all datas under .purple, To support indexing pidgin logs, tracker only index files under $HOME/.purple/logs. > > * How does this work with Trusted Extensions ? for example will there > be a trackerd running per label ? I don't think that there is any difference when run with Trusted Extension. Please correct me if I am wrong. > * Can it be configured to only index certain file types, for example > do my ascii text and OpenDocument files but don't touch my C code. > Sorry, tracker-preference does not provide this option to only index certain file type. but user can choose to ignore some file type in preference settings For example, choose to index home directory, but exclude "*.c" and "*.h", "*.cpp"
