On Sun, Jul 20, 2003 at 06:35:28PM +0200, Timwi wrote:
Tomasz Wegrzanowski wrote:
On Sun, Jul 20, 2003 at 02:36:38PM +0200, Timwi
wrote:
Tomasz Wegrzanowski wrote:
Wasting bytes where it's critical to make RC
as small as possible
for "complicated reasons" doesn't really convince me.
Could you please explain to me why you think it's critical to make RC as
small as possible? Our biggest current problem is database performance,
not bandwidth usage.
Because most people are using 56kbit modems and recent changes size is
huge.
It's not about server bandwidth at all.
A 56kbit modem can transmit 56 Kilobits per second! With TCP's ~3.8%
overhead (let's exaggerate to 4% for simplicity), it means you can
transmit 53.76 Kbit/sec = 6.72 KB/sec. The Recent Changes list as I see
it now (last 50 changes) is 40.8 KB ... this would therefore, under
perfect server performance, take almost exactly 6 seconds.
According to your own calculations, you can save up to 16.6% - that
takes it down to very close to 5 seconds.
I'm not convinced this unnoticeable improvement is worth the hassle.
50 changes is next to useless. 100 is minimum that's any useful, and
250 or 500 is needed to really know what's going on.
250 changes is about 200 kB, or 29s, with 16.6% less to transfer, it's 24s.
Sounds significant enough to me.