Re: metadata question.

Adding it to the in-memory cache index as a temporary workaround is
fine, but the goal MUST be to not have it there.

I don't see that it will need to conflict with ranges. For one thing
partial replies SHOULD have a content-length, for another we need to
handle them differently anyway.

/Henrik

Robert Collins wrote:

> Well, the http code handling reading from the upstream socket
> calculates this (now) on the fly. I'm hesitant to use file size -
> headers, because that _will_ conflict with partial range storage
> down the track. I was thinking that adding it to the swap.state
> metadata would be a workaround - but how much ram overhead will
> that entail - (I wasn't planning to tackle updatable store object
> metadata just yet :-])...
Received on Sun Jan 28 2001 - 15:51:18 MST