someone smarter than Dave Cutler
in [Device Drivers]
|
From: jmfbahciv on 31 Mar 2010 07:21 Jim Stewart wrote: > Mark Crispin wrote: >> On Tue, 30 Mar 2010, Jim Stewart posted: >>> Anyone that takes the time to leaf through some >>> Datamation magazines of that era would be lucky >>> to find any reference to PDP-10's. >> >> Using Datamation as an historical reference is like using the National >> Enquirier. > > A circular religious argument not unexpected from > someone that believed that PDP-10's dominated the > era. > > You, obviously, have had no experience in non-IBM niches. /BAH
From: jmfbahciv on 31 Mar 2010 07:24 Johnny Billquist wrote: > Pat Farrell wrote: >> Jonathan de Boyne Pollard wrote: >>>> _Showstopper_ mentions this. >>>> >>> I'm going to have to read that, someday. >>> >>>> Cutler had to be dragged kicking and screaming into allowing some >>>> paging of the kernel. >>>> >>> I wonder how much of the stick that Dave Cutler gets is completely >>> justified. >> >> IMHO, a lot. He had blinders on the topic of modern memory management. >> The Tenex/Tops-20 folks had it right, and they were all DEC. Now at the >> time, DEC was not really a single company, so a bit of the NIH spirit >> made sense from a corporate view. >> >> But his decisions hurt both VMS and NT, or rather his lack of experience >> in the state of the art. And in some cases, the start of the art we are >> talking about is 1969 art, implemented in the late 70s for VMS and late >> 80s for NT. > > Not ever have been close to Dave Cutler, I can only reflect on some > properties of RSX and VMS, of which I have some experience. > Cutler seems to have been more focused on micro-kernels, which would > lead to a design where you do not want or need paging in the kernel. > > Things which would bloat the kernel, and motivate having a pageable > kernel are outside of the kernel in RSX and VMS. Things like file system > code are in a separate process, for instance. Process context in VMS is > pageable, including even the page table for a process. > > Now, microkernels did come into fashion later on, and is once more very > much hot now. You want to load/unloade modules, drivers and whatnot in a > running kernel, and not configure that statically before even booting. > And the kernel wants to be small (but I haven't seen a single one who > actually is, in my mind). > > Of course, how you design things are always based on what your targets are. > > Cutler was/is definitely not unfamiliar with the wish and requirements > of having a small footprint of a kernel. RSX stands as a very good > testimony to that. > > As far as I can tell, DC seems to be able to rub a lot of people the > wrong way. No denying that. > But giving him the stick for his technical abilities are usually more > because he did things in a different way than the PDP-10 crowd at DEC. > Something they seem to never forgive him for. > But the technical merits in the attacks seldom seem to be very high. > > Sorry, but that's my point of view. And I happen to like TOPS-20 too, in > addition to RSX. And I think there are some things that are very nice > about VMS as well. > Cutler's style was task-based systems, not timesharing. To expect him to think in terms of timesharing is wrong. /BAH
From: jmfbahciv on 31 Mar 2010 07:28 Jim Stewart wrote: > Mark Crispin wrote: >> On Tue, 30 Mar 2010, Jim Stewart posted: >>> Mark Crispin wrote: >>>> On Tue, 30 Mar 2010, Jim Stewart posted: >>>>> Anyone that takes the time to leaf through some >>>>> Datamation magazines of that era would be lucky >>>>> to find any reference to PDP-10's. >>>> Using Datamation as an historical reference is like using the >>>> National Enquirier. >>> A circular religious argument not unexpected from >>> someone that believed that PDP-10's dominated the >>> era. >> >> So enlighten us about the IBM S/360 systems used to build the >> Internet. Tell us how such tools as document processing and email were >> invented on S/360. Tell us about how the world's timesharing was on >> TSS/360, Call-OS, and APL\360. Tell us about the seminal work on >> symbolic algebraric manipulation and artificial intelligence done on >> S/360. >> >> Oh, and tell us how S/360 launched the computer game industry with >> such programs as adventure and zork. >> >> Apparently, in the world of Jim Stewart, everybody used S/360 until >> they switched to Windows. > > As long as you consider "dominated the era" to really > mean "technically superior" and I consider it to mean > "market share", we'll never agree. I consider it to mean numbers of users who had access to computer services/minute. > > Setting that aside, and it's a big set-aside, I question > how much the PDP-10 was responsible for building the > internet. A LOT. > My understanding is that PDP-11's, Vaxen and > IMP's built the early internet. Granted, the PDP-10's > made major contributions to time-sharing and AI, and > they were beautiful machines, but that doesn't mean they > dominated the era. > Sigh! JMF's first task at DEC was to make PDP-10s, PDP-12s, PDP-8s, PDP-11s, IBMs systems talk to each other. That was in 1970 at a site now called ORNL.
From: jmfbahciv on 31 Mar 2010 07:31 Mark Crispin wrote: > On Tue, 30 Mar 2010, Pat Farrell posted: >> Jim Stewart wrote: >>> Setting that aside, and it's a big set-aside, I question >>> how much the PDP-10 was responsible for building the >>> internet. My understanding is that PDP-11's, Vaxen and >>> IMP's built the early internet. > > When it comes to Internet history, Jim Stewart is blowing farts out his > anus and claiming that they are facts. > > I was there in the 1970s. > > Look at > http://en.wikipedia.org/wiki/File:Arpnet-map-march-1977.png > > PDP-10 dominates the picture (note variants, such as "DEC-1080", > "DEC-2050", "MAXC", etc.), especially of systems offering services. > > There were many fewer PDP-11 systems systems than PDP-10 systems on the > ARPAnet. Most were client only systems ("users"). Only a handful of > very large PDP-11 systems (mostly running nascent UNIX) offered any > services. The same is true for many of the other machines listed. I > recall only one IBM S/360 with services, and that was accomplished via a > micro acting as a front end. > > The NIC's maps and host tables were, of course, always horribly > inaccurate. They showed long-defunct machines (e.g., PDP-1) and omitted > new PDP-10 systems which continually popped up. We never bothered > making our own maps, but we did make our own host tables. > >> Vaxen were much later. > > VAX/VMS had a negligible presence on the Internet. > > Almost all Internet VAXen ran BSD UNIX. Their advent wasn't until > 1979-1980. By that time, ARPAnet had been in operation for a decade. > > Most VAXen did not get onto the network until the TCP/IP transition on > January 1, 1983. Prior to that time, if they had connectivity to the > network, it was often over a local area network to a PDP-10 system on > which they could log in and then access ARPAnet. This was the day and > age of "small-i" internet where SMTP relaying was essential as messages > hopped from VAX to PDP-10 to PDP-10 to VAX. > > Even after the TCP/IP transition, VAXen remained second class citizens > for some time due to the miserable PDP-11 based routers. It wasn't > until microprocessor routers, later in the 1980s, that it was reasonable > to be on a non-ARPAnet network and not resort to small-i internet. Even > then, it wasn't until NSFnet in the late 1980s that the ARPAnet finally > began to wither away. > >> Many of the early nodes on the Arparnet/Dapranet were aimed at >> connecting PDP-10 and Tenex systems. The IMPs were minicomputers, or in >> IBM speak, IO controllers. > > IMPs were Honeywell 316 and 516 16-bit computers, and served two > purposes; one as a packet switch to route packets on the ARPANet (over > point-to-point links to other IMPs), and the other being to connect > other computers to the ARPAnet. From the point of view of a computer on > the ARPAnet, the IMP was no more than a hub (although it was very much > more internally). > > ISTR that BBN made a processor that was used as an IMP in the waning > years of the ARPAnet. > >> The very first IMPs were specialized, then PDP-8s, and later PDP-11. > > No. > > PDP-8s were never used as IMPs; nor to my knowledge was a PDP-8 ever > connected to the ARPAnet. > > PDP-11s were (briefly) used as Internet routers before being replaced by > microcomputer based devices (e.g., early cisco routers). They were > never IMPs. > PDP-8s were the first CPU when a user would use to get at the PDP-10. Think about dial-ups and TTYs which were "far away" from the DC-10s. Users didn't see the 8s but those systems were used to answer the phones. /BAH
From: Morten Reistad on 31 Mar 2010 09:16
In article <alpine.OSX.2.00.1003302142210.366(a)hsinghsing.panda.com>, Mark Crispin <mrc(a)panda.com> wrote: >On Tue, 30 Mar 2010, Pat Farrell posted: >> Jim Stewart wrote: >>> Setting that aside, and it's a big set-aside, I question >>> how much the PDP-10 was responsible for building the >>> internet. My understanding is that PDP-11's, Vaxen and >>> IMP's built the early internet. Also, remember that the PDP10's and some '11s succeeded on the nascent Internet very much despite DEC. They ran non-dec software more often than not. Tenex, unix, ITS were popular. >When it comes to Internet history, Jim Stewart is blowing farts out his >anus and claiming that they are facts. > >I was there in the 1970s. I was a little later, on the other side of the pond. 1979 saw a transition from Tops10 to Tops20, and unix came in use. Some odd machines like cybers, primes, etc were also on the net after the tcp/ip transition. >Look at > http://en.wikipedia.org/wiki/File:Arpnet-map-march-1977.png > >PDP-10 dominates the picture (note variants, such as "DEC-1080", >"DEC-2050", "MAXC", etc.), especially of systems offering services. > >There were many fewer PDP-11 systems systems than PDP-10 systems on the >ARPAnet. Most were client only systems ("users"). Only a handful of very >large PDP-11 systems (mostly running nascent UNIX) offered any services. >The same is true for many of the other machines listed. I recall only one >IBM S/360 with services, and that was accomplished via a micro acting as a >front end. > >The NIC's maps and host tables were, of course, always horribly >inaccurate. They showed long-defunct machines (e.g., PDP-1) and omitted >new PDP-10 systems which continually popped up. We never bothered making >our own maps, but we did make our own host tables. > >> Vaxen were much later. > >VAX/VMS had a negligible presence on the Internet. > >Almost all Internet VAXen ran BSD UNIX. Their advent wasn't until >1979-1980. By that time, ARPAnet had been in operation for a decade. > >Most VAXen did not get onto the network until the TCP/IP transition on >January 1, 1983. Prior to that time, if they had connectivity to the >network, it was often over a local area network to a PDP-10 system on >which they could log in and then access ARPAnet. This was the day and age >of "small-i" internet where SMTP relaying was essential as messages hopped >from VAX to PDP-10 to PDP-10 to VAX. > >Even after the TCP/IP transition, VAXen remained second class citizens for >some time due to the miserable PDP-11 based routers. It wasn't until >microprocessor routers, later in the 1980s, that it was reasonable to be >on a non-ARPAnet network and not resort to small-i internet. Even then, >it wasn't until NSFnet in the late 1980s that the ARPAnet finally began to >wither away. We saw better performance from a 3-head QNX (with arcnet) 80x86 machine than from a VAX 785 with unix. Running usenet, conferencing systems, ftp archines, telnet logins etc. This was early 1984. >> Many of the early nodes on the Arparnet/Dapranet were aimed at >> connecting PDP-10 and Tenex systems. The IMPs were minicomputers, or in >> IBM speak, IO controllers. > >IMPs were Honeywell 316 and 516 16-bit computers, and served two purposes; >one as a packet switch to route packets on the ARPANet (over >point-to-point links to other IMPs), and the other being to connect other >computers to the ARPAnet. From the point of view of a computer on the >ARPAnet, the IMP was no more than a hub (although it was very much more >internally). > >ISTR that BBN made a processor that was used as an IMP in the waning years >of the ARPAnet. > >> The very first IMPs were specialized, then PDP-8s, and later PDP-11. > >No. > >PDP-8s were never used as IMPs; nor to my knowledge was a PDP-8 ever >connected to the ARPAnet. > >PDP-11s were (briefly) used as Internet routers before being replaced by >microcomputer based devices (e.g., early cisco routers). They were never >IMPs. Sun's also had a brief career as routers, using VME boards for E1 and T1 lines, and their built-in ethernet cards. I still have some of those VME boards somewhere. -- mrr |