[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: InstantMirror initial source repo

Ed Swierk wrote:
On 11/21/07, Warren Togami <wtogami redhat com> wrote:
This means the server will keep downloading that file multiple times
until the file download is complete, then new client requests that come
in after that point use the cached version?

Yes. We could try to get fancy and have later requests wait on the one
already underway, but it's not clear how you tell the difference
between a file that's still being downloaded and a turd left behind by
an aborted request.


There may be different approaches we can take to achieve what we need, but I am afraid we may have to think about metadata on the side (outside of the mirror tree), and possibly tracking these concurrent demands within the separate daemon. Unless we intend for InstantMirror to be more than just a small office mirror (which would be very handy for many people), we will need to eventually solve this problem.

In any case, great work getting this far so quickly! =)

Warren Togami
wtogami redhat com

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]