[email protected] (Shmuel Metz  , Seymour J.) writes:
> You brought up error recovery. I was hinting that SDLC has error
> recovery and IP doesn't. The error recovery in TCP is at a higher
> level, and I don't see why you expect it to have lower overhead.

part of 80s protocal designs was that modern hardware transmission
technology mvoing to employ sophisticated forwarded error correct (FEC)
... higher level protocols saw a lot fewer transmission errors
... higher level protocols could handle packets lost because of the
fewer packet transmission error ... but also things like packet drops
because of congestion at intermediate nodes.

however, there was also other sort of issues. i've mentioned being
brought in to consult at small client/server startup that wanted to do
payment transactions on their server, they had also invented this
technology called SSL (the result is now frequently called "electronic
commerce") ... and we had to map SSL technology to payment transaction
process.

one of the very early adopters was national sport retailer that was
expected increased web traffic during half-time of sunday professional
football. they had multiple lines into different places into the
internet backbone for availability. one of the complicating factors was
that some number of ISPs still took routers down on sunday for
maintenance (creating outages).

I had originally worked on using routed protocol to advertise
ip-addresses at different routes (for availability)... but the internet
backbone started the process of moving to hierarchical routings ... so
fault masking for availability had to fall back to (DNS) multiple
A-records (instead of host name mapping to single ip-address, it would
have a table of 2 or more ip-addresses).

I explained to the browser group that they had to support multiple
A-records and why (if unable to make connection with 1st ip-address,
continue to try and use any additional ip-addresses) ... they responded
with it was too complex ...  i provided them with example code from 4.3
tahoe/reno clients ... and they still said it was too complex (it took
something like another 12 months before i got them to support multiple
a-records; aka while i had absolute authority of all implementation
deployment for the webserver to payment gateway ... i could only advise
on the browser to webserver implementation).

-- 
virtualization experience starting Jan1968, online at home since Mar1970

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to