I continue to see reports of downloads stalling and then showing dramatic 
improvements after a restart. I'm skeptical. I have not been able to 
reproduce the alleged bug recently: when I leave the node running overnight, 
I come back and I see that downloads are progressing, admittedly slowly. 
However I have a theory:

On restart, every persistent request is started again from 0%. Because almost 
all files are splitfiles, it has to fetch the top of the splitfile first. 
This means it will start at 0% and, assuming it's fetched this part 
successfully before, rapidly progress towards 100% (because it's only 
fetching a few blocks, which are all in the datastore anyway). Then it will 
jump back to where it was, and make much slower progress, because it's now 
fetching the real data.

If I am right, we need to make it a lot more obvious when a percentage is 
provisional (i.e. when it's likely we are going to add more blocks once this 
layer is fetched). We could maybe add a layer/stage indicator, either as a 
number or as a color?

My apologies to all you who have reported this bug, it's entirely possible 
that there is a real download stalling bug, I'm just pondering a possible 
explanation - does it fit with your experience?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20080329/2ebd654e/attachment.pgp>

Reply via email to