Path: csiph.com!eternal-september.org!feeder.eternal-september.org!nntp.eternal-september.org!.POSTED!not-for-mail From: Lynn Wheeler Newsgroups: alt.folklore.computers Subject: Re: 43 Years Of TCP/IP Date: Fri, 02 Jan 2026 13:27:04 -1000 Organization: Wheeler&Wheeler Lines: 81 Message-ID: <87secnppdj.fsf@localhost> References: <10j6l76$3oras$4@dont-email.me> <10j6lpa$3ppd5$1@dont-email.me> <87ms2wkdn8.fsf@localhost> <10j77v7$3a2$1@dont-email.me> MIME-Version: 1.0 Content-Type: text/plain Injection-Date: Fri, 02 Jan 2026 23:27:10 +0000 (UTC) Injection-Info: dont-email.me; posting-host="08af7bd3c1cfa38c38472058c5a61b0e"; logging-data="894805"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX18Xrfg3riEMViYmdvqnvS09w6PKPWLSHp4=" User-Agent: Gnus/5.13 (Gnus v5.13) Cancel-Lock: sha1:v0nm/apQq5eUSMEk3r3zlJGdYpQ= sha1:iV8wnPIIpI4zRnOpmcx7TcCbkC0= Xref: csiph.com alt.folklore.computers:233091 Al Kossow writes: > Chessin came to visit us in the Systems Technology Group at Apple ATG > and we had a nice discussion. > > I had wondered whatever happened to XTP. TCP had minimum 7 packet exchange and XTP defined a reliable transaction with minimum of 3 packet exchange. Issue was that TCP/IP was part of kernel distribution requiring physical media (and typically some expertise for complete system change/upgrade; browsers and webservers were self contained load&go). XTP also defined things like trailer protocol where interface hardware could do CRC as packet flowing through and do the append/check ... helping minimize packet fiddling (as well as other pieces of protocol offloading, Chessin also liked to draw analogies with SGI graphic card process pipelining). Problem was that there were lots of push back (part of claim at the time HTTPS prevailing over IPSEC) for any kernel change prereq. topic drift ... 1988, HA/6000 was approved, initially for NYTimes to migrate their newspaper system off DEC VAXCluster to RS/6000. I rename it HA/CMP https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing when I start doing technical/scientific cluster scale-up with national labs (LANL, LLNL, NCAR, etc, also porting LLNL LINCS and NCAR filesystems to HA/CMP) and commercial cluster scale-up with RDBMS vendors (Oracle, Sybase, Ingres, Informix) that had VAXCluster support in same source base with unix (also do DLM supporting VAXCluster semantics). Early Jan92, have a meeting with Oracle CEO where IBM AWD executive Hester tells Ellison that we would have 16-system clusters by mid92 and 128-system clusters by ye92. Mid Jan92, convince IBM FSD to bid HA/CMP for gov. supercomputers. Late Jan92, cluster scale-up is transferred for announce as IBM Supercomputer (for technical/scientific *ONLY*) and we are told we can't do clusters with anything that involve more than four systems (we leave IBM a few months later). Partially blamed FSD going up to the IBM Kingston supercomputer group to tell them they were adopting HA/CMP for gov. bids (of course somebody was going to have to do it eventually). A couple weeks later, 17feb1992, Computerworld news ... IBM establishes laboratory to develop parallel systems (pg8) https://archive.org/details/sim_computerworld_1992-02-17_26_7 Not long after leaving IBM, was brought in as consulatnt to small client/server startup, two former Oracle people (that had worked on HA/CMP and were in the Ellison/Hester meeting) are there responsible for something call "commerce server" and they want to do payment transactions. The startup had also invented this stuff they called "SSL" they want to use, it is now frequently calle "e-commerce". I had responsibility between web servers and payment networks, including the payment gateways. One of the problems with HTTP&HTTPS were transactions built on top of TCP ... implementation that sort of assumed long lived sessions (made it easier to install on top kernel TCP/IP protocol stack). As webserver workload ramped up, web servers were starting to spend 95+% of CPU running FINWAIT list. NETSCAPE was increasing number of servers and trying to spread the workload. Eventually NETSCAPE installs a large multiprocessor server from SEQUENT (that had also redone DYNIX FINWAIT processing to eliminate that non-linear increase in CPU overhead). XTP had provided for piggy-back transaction processing to keep packet exchange overhead to minimum ... and I showed HTTPS over XTP in the minimum 3-packet exchange (existing HTTPS had to 1st establish TCP session, then establish HTTPS, then the transaction, then shutdown session). https://en.wikipedia.org/wiki/Xpress_Transport_Protocol other trivia: I then did a talk on "Why Internet Isn't Business Critical Dataprocessing" based on documentation, processes and software I had to do for e-commerce, which (IETF RFC editor) Postel sponsored at ISI/USC. more trivia: when 1st started doing TCP/IP over high-speed satellite links, established dynamic adaptive rate-based pacing implementation ... which I also got written into the XTP spec. -- virtualization experience starting Jan1968, online at home since Mar1970