Re: 28.8 Upgrade

Michael Sattler (msattler@jungle.com)
Tue, 6 Jun 1995 16:03:41 -0700


At 4:55 6/6/95, Christopher M Sedore wrote:
>On Mon, 5 Jun 1995, Michael Sattler wrote:
>
>> Disagree. The quality of the line you get with a modem has nothing to do
>> with your provider, but with your local phone company.
>
>Both, wouldn't you say? If your provider has a 56k link and 15 people
>competing for it, that line quality (perhaps better stated "capacity"), in
>addition to your phone line, will be a major determining factor. Same
>goes for a net provider who has poor connectivity on his providers
>net-I've seen T1s into the net that only allowed 5-7KB/s to the NSF backbone
>because the net provider's backbone net was severely oversubscribed.

True. I was assuming a moderately ethical ISP - someone that
oversubscribed is selling the idea of net access rather than net access
itself.

>> Something that's been unmentioned here is that PPP takes an extra five
>> percent bandwidth away from you for it to be able to support multiple
>> protocols. SLIP only does IP. That 5% should be enough for modem users to
>> demand SLIP in bandwidth-intensive situations such as CU-SeeMe.
>
>I'd like to see how you come to those calculations. The PPP difference
>should be something like 5 bytes per packet more than SLIP, which works
>out to little more than .5% on a 1000 byte mtu or 1.9% on a 256 byte mtu.

I didn't do the math, it came directly from the people who wrote MacTCP and
the team that is doing Open Transport. I believe it's more than just the
packet size, but the CPU load of checking each packet to see what protocol
it is.

-----------------------------------------------------------------------+
Michael Sattler <msattler@jungle.com> http://www.jungle.com/msattler/ |
|
To leave this list send listproc@cornell.edu a message body with the |
command "u*subscribe cu-seeme-l". (Replace the '*' with 'n'; if |
the "un" word had appeared then this message would have automagically |
been filtered to the trash by many people's filtering configuration.) |