Matthew Toseland wrote:
On Thu, Sep 01, 2005 at 10:17:51AM +0200, Alex R. Mosteo wrote:
Given that all the downloading/assembling is done inside the node, you
just give to the browser a standard download connection[*], and make
available the data that is available contiguous from the start of the
file. This surely means that the download will go in jumps when holes
are filled, and it may seem stalled for some time, but at least you have
a regular download whichever the browser.
You do, however, the user will assume it is stalled and cancel it, since
it won't return any data at all for minutes to hours on end.
Well, it's a matter of advertising that freenet downloads can seem
stalled for long times, just like now a long time for node integration
is expected for new nodes.
But I wonder if there are browser-side timeouts.
The API for 3rd party apps allows for more detailed download
managers/multi-node downloads for those entrepreneur developers.
I really think that, even if some quantity of rounding squares is
needed, providing a standard download is in the benefit of freenet.
However, a "download feedback page" is too a deviation from standard
behavior, and I agree that it should not be repeated.
Giving a standard download connection would allow too for cancelling
downloads in the usual way.
And they would *always* be cancelled. The ONLY way to make this work
would be to download the data in such a way that it works - i.e. to
choose the random blocks first, then download the data. The problem with
this is that then everyone will be able to get the first few megs of the
file, but later parts will fall out. We need all parts of the file to
drop out of the cache at the same rate. That means fetching in random
order. That means we could well have most of the download before we have
much sequential data - especially if we use the onion codec. It's fine
for small files, where we can fetch every block at the same time. But
it's not fine for large files.
Depending on the circumstances, I could even have some time to implement
this if feasible and nobody is interested in future months.
[*] I just mean answering to the request with a standard HTTP reply, and
provide the sequential data when it's available.
-- jeek
------------------------------------------------------------------------
_______________________________________________
Devl mailing list
[email protected]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
_______________________________________________
Devl mailing list
[email protected]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
