The gtf format is one of the worst, but because of tophat it will be used a lot.
Illumina are now providing some (also for arabidopsis):
http://tophat.cbcb.umd.edu/igenomes.html
Whats nice in IGB is the split by protein-coding, pseudogene etc.., whats bad is that the exons are not connected. I tried it with c elegans from ensembl.
Please check if the display is as it should be.
I second this. We should support tophat and cufflinks gtf.
Changing the topic slightly: Why aren't tophat and cufflinks using bed? If no-one can think of a reason why that format wouldn't work, maybe we can request modifying the software to allow it. TopHat produces junctions.bed already.
We're going to be seeing a lot of these files shortly so if possible I'd request building their distribution into the DAS/2 server. Ditto for xxx.var files.