|
Re: example .procmailrc stuff for announce lists
2001-11-23 11:20:02
On 11/23/01 at 9:28 AM +0100, Harald Alvestrand wrote:
Perhaps 1000 mail messages per year, each containing about 6
KBytes, is a significant extra load for Mr. Resnick's local network
and machine. Since 6 MByte in a year is equivalent to at most
about 150 HTTP pages in a year, Mr. Resnick's network and machine
must not be used for "web surfing." Is that true?
real numbers are nice.
it's approximately 4 kbytes, and up to 50 messages per day during
"hectic time" (such as November 16). I think the total is around
4000 messages/year.
so at 33.6 kbits/sec, it is around 45 seconds of added download time
for each day of I-D announcements. Your modem speed and efficiency
may vary.
Ah, if "real" numbers were only real. First let's discuss the actual
setup that's causing the problem (and to an extent Vernon is correct
in his assessment): It's not my regular DSL network connection which
I use when I'm at home, but my connection back to that network when I
am using my Handspring Visor on the road. I've got a maximum 14.4kbps
connection over my mobile phone back to my home network. So, even if
life were perfect, on the hectic days of 50 messages at 4K a message
(200kB or 1.6Mb of data), that's about 2 minutes of download time per
day. But as we all know, life ain't perfect. First of all, on those
hectic days, all of those messages come down at once in my morning
download, so it's not evenly distributed over the day. And given
network slowdowns on a wireless link from time to time, the extra
data produced by the actual protocol work, the round trips and
latencies involved (especially if I choose to do IMAP instead of
POP), and processing for the filtering, we're actually talking about
more like 4 to 6 minutes of online time just devoted to these
messages. In the course of a life, not a big amount of time, but I
would claim that 4 to 6 minutes added to my mail check is a pretty
serious hit. (It would solve my problem I could do server side
filtering. For assorted reasons, I can't.)
So, issue number one for me is the above scenario. But the other
interesting question is the hit the secretariat takes. Even if there
were only 1000 subscribers to IETF-Announce, that's 200MB of data on
the busy days. However, my guess (and the secretariat can give us the
real numbers) is that that number is at least an order of magnitude
too low. My theory is that a serious chunk could be taken out of the
overall output if I-D's were separated out because not nearly as many
people would subscribe to get all of the I-D announcements. Given the
low investment of time it would take to test this out, I think it
would be worth attempting it.
pr
--
Pete Resnick <mailto:presnick(_at_)qualcomm(_dot_)com>
QUALCOMM Incorporated
|
|