Hi snp_analyser. I'm afraid there is no good value  the inner distances depend on the sizes of your fragments and the lengths of your reads, which are experiment specific.
We typically use a tool like Bowtie to help us find our fragment sizes empirically. We run a pairedend alignment with Bowtie, using default parameters for I and X. We then examine the output to see, in general, how far apart the reads in a pair as aligned. This indicates the mate inner distance.
In terms of terminology, the "gap" or "inner distance" is the distance between the aligned reads (not counting the reads themselves). The "insert size", on the other hand, includes the reads themselves, so can be thought of as the "fragment size".
If you look at Bowtie output, the alignment position of a read is the position of the first base in the read (from the perspective of the forward reference strand). This means that if you subtract the alignment positions of the two reads in a pair, the result is actually "inner distance" + "read length". So you will need to subtract the read length to get the inner distance.
You should probably write (or find) a script to do this for you to ensure you examine enough pairs to get a representative feel for the value. Our data typically shows a normally distributed inner distance.
