My Photo

Photo Albums

Blog powered by Typepad
Member since 02/2004
AddThis Social Bookmark Button

« A National Disgrace II | Main | 1970 »

Monday, 14 February 2005


Celestial Weasel

The hype over the Cell reminds me of the hype over.. um.. let me see, Transmeta, RISC, the transputer (to name but 3 processors). Frankly the chances of Sony and IBM coming up with anything that takes over the world are about nil, basically any joint venture with IBM turns to s**t and this is not the sort of innovation that Japanese companies do. What has come out of Japan in the way of computer tech? MSX and the Ruby programming language and er....

Incidentally, if you read the Register, do you think that Andrew Orlowski has 'lost it'? He has been in California too long, if you ask me.

I must admit to knowing very little about the Cell, life it too short to read very much about things that don't look like they have a hope in hell of succeeding.

Also, the worst service anyone can do to any hardware is to port Linux to it - the great lowest common denominator OS that elimates any advantage any hardware has over the PDP-11 hardware it was written for. OK, I exaggerate slightly for effect.


The Cell is going to power the PlayStation 3, so, assuming that the Xbox 2 doesn't capture most of the next generation console market, the Cell is likely to ship in volume. Now, it's true that the PS2 Emotion Engine was touted as a general purpose chip and this never happened. But the Cell is made by IBM, who have great success with the POWER processors at the moment and they have every reason to push the Cell. Furthermore, the Cell is much more powerful and flexible than the increasingly tired looking x86 series. The Cell could catch on where Transmeta and others failed. It's not going to vapourware (although that doesn't mean it will live up to its hype) and it does offer significant performance advantages over other chips (wasn't Transmeta a low power chip?). As Unix, well, I take your point. Perhaps we should port VMS ;)

I stopped reading The Reg regularly a couple of years ago when it became dull. Orlowski has certainly lost. Being half of Belle de Jour with his girlfriend Sarah Champion probably hasn't helped his state of mind (I feel on the balance of probabilities that a tech journalist living in San Francisco is more likely to have heard of and refer to the Singularity than a call girl, even one with a literary bent, in London). Where you thinking of any specific articles or his comments about open networks in his posting from Cannes?

Celestial Weasel

Although Andrew O. is a welcome counterpoint to the Boing Boing manics, he seems to have gone somewhat over the top with his anti-Google anti-Internet hobby-horse.

I guess, with the PS and PS2, the MIPS processor ships in volume, too (does the PS2 have a MIPS in it?).

I always thought that the ARM was a good processor. I remember many many years ago (late 80s) seeing a book on programming it in Assembler in the 'good' bookshop in Cambridge, Hatchards is it? (this is back in the days when there were about 4 good bookshops in England). Perhaps I should have bought it.


Heffer's. Owned by Blackwell's now (I think). Yes, those were the days when a visit to Dillon's in Gower Street was something special.

Orlowski is also very anti-blog, which is one reason why he is (half of) the prime suspect to be BdeJ (the suggestion is that it was set up as way of discrediting blogs).

Celestial Weasel

I've probably said it before, but whilst it is wonderful to be able to get hold of any in print and many out of print books, as well as all the info on teh interweb anyway, but there was a certain thrill in going to Foyles / Dillons / Compendium / The Libertarian Bookshop / Forbidden Planet which is now lost.

Yes, Heffer's of course.

I haven't ever read BdJ, it kind of falls out of my area of sociological interest - why anyone with a functioning left / right hand would pay for sex is completely beyond me. I also have never been able to get very excited over these 'who is the real BdJ / Deep Throat / Anonymous author of Primary Colours / Chris Moyles' stories (and who IS this Paul guy anyway?)
What little I have read suggests strongly that BdJ isn't 'for real', though I have difficultly imagining Mr O. doing this, whilst he is quite wry at times, this seems somewhat too frivolous a venture for him.
The 'singularity' thing does seem a slightly bizarre slip by the author, though it is good to see you mention the S word, you have gone a while without using it and I was beginning to worry.

In terms of anonymity, what would be good would be for an OS of simplicity and elegance (maybe for the Grid (!)) to appear from nowhere (obviously it would have to be for some constrained hardware). Then we could all speculate as to whether it was a leak from a company or a 'Rocky Chipset' working alone (in one of his few sane posts, Winer said something on these lines, i.e. it would be good for some software to appear out of nowhere by an unknown genius - the name Rocky Chipset is by analogy with Tiger Woods.


Well, I suppose Linux initially appeared out of nowhere from an unknown genius.

A problem with writing an OS from scratch is that it would have to at least attempt to be Posix (or rather these days SUS) compliant, which raises the question of why not use just use Unix (OK, I know Linux is not strictly SUS-compliant, but it's close enough for government work). I remember when I was at Ericsson, we managed (somehow) to persuade the system administrator to install the NT Posix package (there was some good reason why we wanted it, which I have now forgotten). One of my colleagues was amazed that such a thing existed at all. I was surprised that it available and actually worked.

Celestial Weasel

No, no, no. On the lines of 'the revolution will not be televised' I propose 'the singularity with not be Posix compliant'.

Have we had this discussion before, I forget?

Absolutely everything should be up for grabs. Anything you like could potentially be added as a compatability layer on top (e.g. a Posix layer on top). Maybe it doesn't allocate memory or disk in the 'C' style 'bucket'o'bits' manner. Maybe it doesn't have the concept of files, or directories.

And as for Linux, reimplementing Unix certainly doesn't count.


There are people who believe that not only is Unix the best operating system there is, it is the best operating system there could be.

These people are wrong on both counts.

I'm hoping that whatever OS we use after the Singularity has a undelete command. Or at least multiple file versions. And no rm -rf *

Celestial Weasel

And no file globbing with amusing side effects if you have a file called -rf in a directory...

You are quire right, it is quite clear that the C / Unix 'bucket'o'bytes' paradigm is so dominant that people are genuinely unable to see that it is not the only way that things could be done.

'VMS - the only operating system that will get you past the Singlularity' (TM)


VMS: God's Own OS (as Dr Ayde refers to it).

The bucket o' bytes approach has advantages and disadvantages. What gets is the internal inconsistency of the whole thing. I have a book on Unix (quite a good one actually) that tries to teach Unix through the Unix philiosophy (bucket o' bytes). But really it is more a case of "Let a Thousand Philosophies Bloom". For instance, why do the flags for tar not require a "-" before them? I know it can be argued that it is "just one of those things", that life is full of "those things", that the guy who wrote the original tar utility didn't like "-"s before his flags, that the strength of the Unix philiosophy is that it allows this kind of thing with falling over. But personally I'd prefer it if all command flags were handled in a consistent way. And tar should be called something more obvious like "archive". Yes, Unix may support a more heterogeneous culture of utilities that VMS, but unfortunately, this does mean that Unix degrades gracefully, rather the opposite.

Celestial Weasel

Ha, I have just managed to lose my witty reply. I never really got into VMS, though I used it a bit. I guess LispOS and VMS will have to fight it out for the title of God's own OS.

Presumably you have read the Unix Hater's Handbook?

Incidentally, I see you have a link to the bastardised Jargon file, have you seen the original pre-Raymondised version?


I'm going to a future entry on Things I want to Learn or Find Out More About. One of the things I want to learn is Lisp. I even went as far as to download Cygnix to my PC last weekend. Lisp represents a very different approach to computing than the Fortran/JCL - C/Unix one and one I'd like to find out more abou. It's interesting that though RMS began as a Lisp hacker at the MIT AI lab and was driven to set up the FSF by the prospect of Lisp machines running non-free software, GNU, although it may not *be* Unix, is certainly something very *like* Unix.

I have a soft spot for ESR's Jargon File (I have the book). ESR has done some worthwhile things, even if he has many opinions with which I would agree. I will have a look at the pre-Raymondised version. The authentic voice of the MIT AI Lab and SAIL.

I have never read "The Unix Hater's Handbook" although I would give my right testicle to do so. Aeleen Frisch raves about it in "Essential System Administration", but it has been out of print for years. It is available on the web, but it ia huge PDF that would not print properly on the printer at Siemens (memory issue or corrupt file?). Perhaps I should try again, but print-outs are free for me at Imperial!

Celestial Weasel

I have a copy of the original if you want to borrow it. It is still on the web.

There is an article on the Cell as the latest Cringely. His view is not dissimilar to mine
although clearly at some point it IS going to be worthwhile to make the leap..

It has seemed fairly clear to me from a long thread somewhere that Stallman did not really 'get' Lisp, i.e. he was using it more as 'C with brackets'. Not that 'C with brackets' would be a bad thing per se. I will dig up the thread.

By the way, are you going to reply to the fine comment by Rose in another thread?!


I do intend to reply to "Rose", but I feel that it is best to make a response a *considered* one (or at least one I have slept on).

I'm certainly interested in your thoughts about Stallman, Lisp and C with Brackets.

Celestial Weasel

On Stallman and Lisp, here is a thread of emails (obviously there is a possibility that they have been selectively quoted).

I have essentially inferred from this that Stallman was basically using large arrays of things to pass parameters around without bothering too much about abstraction etc.

I also never understood why Stallman, being familiar with Lisp / Lisp machines and Unix decided to clone Unix.

The essential concept of 'C/C++ with brackets' would be to have a language that had the same characteristics as C/C++ e.g. a compiled early-bound language but with a more uniform syntax, thus enabling one to parse the code more easily to perform transformations on code - the fact that Lisp has this is one of the things that one can choose to like about Lisp, but is to my mind pretty much orthogonal to some of the other things one could choose to like about it.
Certainly with our 1 million lines plus of C++ I would rather like to be able to perform such transformations - search and replace is a very poor substitute.

The comments to this entry are closed.

Books We are Reading