Magnus Sundberg wrote:
Maybe, but I beleive that the mail clients are quite simpleminded.
Please broaden my view and give me more examples of when the client
requests unusual headers.
I beleive it requests the headers when it displays the content of the
mailbox. When else does it request the headers?
Mail filtering?
I'm not sure, but I think filtering and things that look at specific
headers (spam tags etc..) do so by downloading the entire message and
then checking, so that isn't it. I'm thinking of things like evolution
when I do an advanced query, but upon further inspection, there are only
about 10 fields (headers) that I can choose from so that isn't such a
big list.
Yes you are right, it should be configurable which headers to always
cache, which to never cache and a threshold level.
The sweet spot in this question, at what level do we create the lowest
database load?
We could create something like a moving average on header requests.
<semi-relevant comment>Try to predict user patterns! You will almost
always be wrong. Instead try to create a system that adapts to the
user pattern, and you would get snappier response.
Compare to the linux interactivity scheduler
</semi-relevant comment>
Sounds good. Yes finding the sweet spot is the key. Some type of
adaptive system would be neat, but step one should be getting the basic
mechanics working, adding the adaptive stuff should come later. I would
hate to see the whold concept not get implemented, or even just take
longer to get implemented because we are waiting on sophicsticated
features. I would be very happy with just an admin configurable list of
cached headers, anything beyond that is gravy.
