in this link http://www.ncbi.nlm.nih.gov/Sitemap/samplerecord.html
in the first CDS we find /translation="SSIYNGIS..."
I tried it to find this translated protein and I found that we can get it if we read 5'3' 3rd reading frame from the origin sequence.
**is that true ? and why did they put the origin from the 3rd frame reading ?**
The first line of the feature to which you refer looks like this:
CDS <1..206
This means that the complete DNA sequence of the gene product is not contained within the sequence defined by ORIGIN.
So how do the annotators know that translation should be in the 3rd frame? Most likely because when they translate the entire DNA sequence in all 6 frames, the product in the 3rd frame "looks like" a real gene product: i.e. it is in-frame (no stop codons until the end at 206) and has similarity to known gene products.