51

Re: Re: when the false step has been made? (A sight from 1983)

Hello, Privalov, you wrote: P> And in case of total  it is necessary to pass to "Mir-3", the Analyst and R-technology. Instead of it is terrible, I when on  I LIE studied, since the second year had half-rates of the laboratorian (and a certain share of spirit) for support of two -1 and -2. Also repaired, and ... And the diploma I on FROM without the Analyst a horse-radish would write the (without calculation errors). Then even article in  quitted.

52

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pestis, you wrote: P> was not. The more widely the bus, the is more than logical elements and the the decision is more expensive. And 100 eight-bit machines would be more useful to the USSR than 10 . And actually it is unique  from all these variants there could be one, 48-digit. The core of applications of a computer at the moment of decision-making on copying IBM-360 is scientific computations. Optimal long words for them it just 48 (either 45, or 46) discharges.  a computer of that time with such long words. Yes, they RISC, somewhat, are not able to process the information byte by byte, and happened the doge have no commands for arithmetical operations with integer numbers. Digit capacity reduction would lead to reduction of high-speed performance by orders. P> also what hinders, except ? The machine on 16 Z80 I saw the eyes. Also what, it could replace at least partly one BESM-6 developed for 20, and can and for 30 years earlier? P> That's it, because architecture of these machines initially defective. And here in a computer initially designed under the multiprocessor architecture it is possible to receive almost linear growth of productivity. It will cost only in 1000 times more expensively single-processor, to break time and again in day, and every minute, and hardware resources on join in system it will be necessary more, than in all processors together taken. It under condition of usage of element basis accessible in second half 60.... <<RSDN@Home 1.2.0 alpha 5 rev. 1495>>

53

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pagid, you wrote: P> And actually it is unique  from all these variants there could be one, 48-digit. P> the core of applications of a computer at the moment of decision-making on copying IBM-360 is scientific computations. Optimal long words for them it just 48 (either 45, or 46) discharges. , from what such output?

54

Re: Re: when the false step has been made? (A sight from 1983)

Hello, netch80, you wrote: N> Hmm, from what such output? That that the main scientific computations? Or long words?... <<RSDN@Home 1.2.0 alpha 5 rev. 1495>>

55

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pestis, you wrote: P> From the point of view "Sciences and Lives" all so. And from the point of view of practising programmers, there is enough enough easily achievable limit of complexity of the project after which it collapses under a body weight. For this reason all necessary software  for a long time rewrote from the Fortran with Kobolom on the modern languages. I at all do not represent that now "actively saw" on the Fortran if all serious calculations do today on GPU, and ancient  the code on GPU is simple so not . It is Just better , than, for example, the code on a C. Because when to the compiler speak "to increase matrix A by matrix B", he in a state to calculate, as it is better to do this multiplication, where  how to organize data streams, etc. And if it has a scanning of the same in a cycle - it still should contract in the beginning to understand that it . And even the C ++ with object "matrix" here does not help - if not to enter a layer of transformation with the multisequencing, depending already from target architecture. That is level of the same compiler. Partially it dares means of type OpenMP, but also at them the typical decision it "to launch 2*K ". For this reason in the Fortran since 80 do at compiler level matrix multiplication,  vectors and matrixes with internal step (! Is, for example, a submatrix from all odd lines and even columns) and so forth. P>> however, these "unsuitable"  for some reason implemented. P> implemented. Perdjachim steam and cheerful dancing on a rake within years 15. Or you forgot problems with incompatibility  which industry solved all 90? It is possible to think, now better. Only this incompatibility was moved on a present first line - GPU, Javascript and so on. P>> And how it was necessary? At once to do the 128-bit address for TCP/IP? P> the variable length Address, from 8 bits indefinitely. And generally all problem there because of that that any fan  bits mixed addressing and routing in one field. At first, any project dividing addressing and routing, did not fly up till now above laboratory level. And it is no wonder - even at units of the size BGP already the half-million of receptions, and will be more further and if to consider that for sufficient speed they should be stored in routers in fast static storage together with other parameters (it is considered, cost to increase for 20) the price of such sharing becomes more than very heavy. Everyones mobile id not to recall is a decision of the same plan, as DNS - simply layer . Secondly - authors of similar ideas somehow are very quickly blown off, when it suggest to draw, as there will be a fast decision at level ASIC. You heard something about X.25? There just the variable length address. And handling at great speed appeared very heavy on what it died, remained transport of banking networks (at the best), and now and there dying out. Now could work the decision in the spirit of - either 32 bits, or 64, and 64 forms by duplication of IP-heading after the first copy. It yes, still is normally laid down on top level routers. But authors IP6 preferred to arrive more considerably, saving convenience to iron. Their decision is not pleasant to me, but I at least understand, why it such. P>> and how it was necessary? What software thought over from the point of view of architecture? P> a software of the end of 90 beginnings 2000. Compare the same Subversion and RCS, bash and sh, nano and ed. Subversion -  states "started to lead to the normal type, all old good broke, new did not finish". At what here it? If to recall good, already Git, Hg. . (Between two last it is possible to argue to a hoarseness, but both of them in times it is better SVN) And yes, RCS it is better - that fairly fulfills the minimum of functions and does not claim for another. nano and ed - as it is possible to compare full screen and a line editor? And by the way, nano  shit, I everywhere take down it and I put joe, the same style, and and possibilities in times is more than convenience. (Though itself I prefer vim, even for letters.) bash from sh - there are not enough differences. Still examples? Meanwhile at you all any doubtful. P>> and what alternative? P> decentralization, the distributed calculations, multitasking and parallelism. Under the total to that also came, now consider on clusters from cheap  gland, instead of on specialized computers with the processor in size with a room. Easier because to increase capacity of one processor further it is impossible. The animal of format Xeon E7 already "caves in" under own gravity. It I do not recall an output suitable, which, on hearings, there the lowest. And about " iron" fairy tales is not necessary, take  top level (as the same Xeon), instead of collect a cluster from , last was unprofitable even on those models, which else gathered in SMP.

56

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pestis, you wrote: P>> And RISC' as the formulated concept then was not. P> titles were not, and the concept was always. No. Concept RISC arose when started to think 1) about extreme cheap implementation of parallel execution of commands 2) about __ orientation to code generation by the compiler, instead of the programmer, with the registration of specificity of compilers. On the first point - classical RISC it when a command 1 at 5th stage, 2 on 4th, and so on while 5th is only just selected from storage. Computers of an epoch to S/360 basically did not assume possibility of execution more than one command, there and one to be in time. On the second - RISC could develop only when there was a possibility to the compiler to be on the average cleverer than the person. Before correspondence of any  an exhaust on the assembler accelerated it in times if not on orders, and it was considered as the normal; And architecture under it were sharpened (look in VAX at that command CALLG does - now for such expenditure of resources would contract a head to the one who offered it). RISC as it stands, speaking to language of the Soviet philosophy, dialectically leaned on limiting CISC, denying it. But on those capacities such it was impossible. Now it is possible to say that any architecture of those times it actually RISC (the same S/360 in the main dial-up of arithmetic-logical commands is qualitative to a floor-RISC), but then it so was not called and it was not realized. P>> you suggest to create and develop any one architecture, so and apparently microprocessor though at the moment of decision-making on copying anybody and did not reflect on them. P> on what element basis to do the processor it is a question absolutely the tenth. Though on transistors, though on lamps.  architecture allowed to take out from  a processor cabinet on  to replace with its board on the discrete logic with the coordination circuit saving already debugged , needlessly to build new date-center from zero. And then to replace the discrete logic with one chip. Partially and was. But for this purpose it was necessary, that the architecture was competitive. Instead of as died in the beginning 90, for example, VAX and M68k. And even z/Arch which the application-oriented code from S/360 fulfills till now without emulation, sits on the brink of leaving in completely  a product.

57

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pagid, you wrote: P> Hello, netch80, you wrote: N>> Hmm, from what such output? P> that that the main scientific computations? Or long words? About a utility of length of 45-48 bits.

58

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pagid, you wrote: Q>> About the atomic I do not know, and in rocket hardly. -1 and could not launch, even in spite of the fact that it conceded to "Saturn" on load-carrying capacity. P> unless it is military projects? P> Americans spent for the lunar program in times more. And at us blinded -1 from this that was. Well here you see - means they were not up to standard of the best. Q>> here therefore was not so because samples for copying could not be received. P> the difference was that for military branches fundamental things - a science were put in expensive and giving return in many years, . Yes anybody was not put anywhere!  always uses that is already invented and it is opened.

59

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pestis, you wrote: P> It is possible, why it is impossible. P> 1. Programming languages. With rare exception type of a C and Lisp from the modern point of view it is a pitch gloom. Algol, Kobol, the BASIC, the Hell, the Fortran Algol it, very roughly telling, the same Pascal. Which is now quite live. A hell it is close to it is also its standardization on disposal of curve ghost effects it is extremely useful in practice.  - it agree, a gloom. A BASIC - quite decent language which has suffered only for an esthetics and reputation. The Fortran - yes, for the present world is syntactically inadequate, though it is possible to adapt. And it for a long time already general purpose language, including even OOP. So from 5 points it agree only on one. P> 2. Standards. Almost all of them , , are unreasoned and unsuitable for . It is Enough to recall ASN1, teletypes, a G-code and RFC epoch of "paper letters". Teletypes here at what? I still somehow agree in zoom-in on ASN.1 though now not problem simply to take library and to draw pair of files of syntax. About RFC somehow did not understand. P> 3. A software. All inconvenient and unreasoned from the point of view of architecture. It is enough to recall RCS, P-CAD, M4, sh, ed and generally unix. It if inside not to look and if to glance there at the best structured programming, and noodles from goto are faster. And what claims to unix? And what to ed, with the registration of conditions under which it formed? P> 4. Concepts centralized , mainframes, supercomputers, architecture it is horror flying. P> 5. The approach to project management naive, one only printout of source codes on a paper for archive, without speaking already about  in log. So it could not be better than anything with level of that time.

60

Re: Re: when the false step has been made? (A sight from 1983)

Hello, netch80, you wrote: N> About a utility of length of 45-48 bits. Not 45-48. From 36 in 50 at IBM-7xx, 45-47 at Soviet -20/220, 48 at BESM-6 and CDC-1604, and 60 in CDC-6xxx For the middle 60 48-digit word was norm. Probably seven decimal significant figures it is poorly, 16 decimal meaning (64-digit) for that element basis was an excess. In the seventieth norm there were 64-digit floating-point numbers.... <<RSDN@Home 1.2.0 alpha 5 rev. 1495>>

61

Re: Re: when the false step has been made? (A sight from 1983)

Hello, netch80, you wrote: N> Kobol - agrees, a gloom.  it is too verbose. And so bearded uncles on it wrote system utilities. That with the Assembler not to communicate. N> a BASIC - quite decent language which has suffered only for an esthetics and reputation. For the sake of justice, early versions of the BASIC is really a gloom. For example, the version used in "Spark-226" is that, what even Hell language looks in comparison with it something paradise. All variables - global. Names can consist no more, than of two characters, and the second - mandatory digit. , the array and a scalar variable can have the same name. In array declaration parentheses were used. Here everyones there TurboBasic, MS Basic, even Visual Basic 6 - already normal.

62

Re: Re: when the false step has been made? (A sight from 1983)

Hello, Glory, you wrote: Hello, pagid, you wrote: P>> Hello, Glory, you wrote:>> And then to copy -52, behind inability to make. P>> also what is the Soviet copy -52? I feel a smell of the expert in types . Sorts of nuts I do not understand. If you do not understand, do not climb in talk.

63

Re: Re: when the false step has been made? (A sight from 1983)

Hello, , you wrote: > It is thought, what exactly error here and was. After mastering of the first variant of foreign system and, probably, its partial adaptation was not any necessity to continue adaptation, using all new and new variants of a prototype. Hardly these variants carried with themselves something in essence new that we could not make on own basis, by own strength and with smaller expenses. And what, it is absolutely difficult to understand, what our lag is caused not by architecture, and integration scale of chips?

64

Re: Re: when the false step has been made? (A sight from 1983)

Hello, Privalov, you wrote: P> Kobol is too verbose. And so bearded uncles on it wrote system utilities. That with the Assembler not to communicate. P> here everyones there TurboBasic, MS Basic, even Visual Basic 6 - already normal.  "normal" it when at you is links/indexes. To verboseness almost you do not pay attention when to you it is necessary to do the bound list/tree through arrays. By this criterion the normal BASIC appeared only in.NET. About  it is not assured, it is possible it till now is not present.

65

Re: Re: when the false step has been made? (A sight from 1983)

Hello, novitk, you wrote: N>  "normal" it when at you is links/indexes. In PL/1 there are pointers. But to name its normal...

66

Re: Re: when the false step has been made? (A sight from 1983)

Hello, , you wrote: > - the Clerk? - Sang, being restricted to arguing, , in a sheepskin coat from hare > the furs, covered I become blue . - I will let know the clerk! Who does it speak - the clerk? (Gogol, Night before Christmas) PL/1 incorporated all rake from the Fortran, Kobola, probably, Algol. Also added a little the.

67

Re: Re: when the false step has been made? (A sight from 1983)

Hello, , you wrote: > PL' they too different happen, There is also good I did not see such. And that saw, looked here is how: > In the middle of 1978 to me doctor Frajburghaus from corporation Data General allocated nearby to Boston addressed. He tried to simplify PL/I firms IBM, a dinosaur made absolutely as if it Disney was engaged. > please, understand that I had to live with this silly, committee developed, language IBM PL/I on IBM System/370 within three years under the threat of conscription. That was worst of ever developed languages, conceding in it only to a Basic language though and it is not enough. In my opinion, state PL/1 is characterized approximately so: After "issue" PL/1 has been certain by some experts as the major language of 70th years. However soon became obvious that he not in a state to "struggle" with  and the FORTRAN, and then, in the late seventies, gave way to the BASIC and PASCAL. http://notes.sochi.org.ru/1164/ I, truth, do not know, in what state there was mentioned BASIC and Pascal at that time. But here the Fortran and Kobol anywhere did not get to.

68

Re: Re: when the false step has been made? (A sight from 1983)

Hello, , you wrote: > And PL/1 anywhere did not get to (for example, in most IBM) About inside IBM I know. And out of IBM itself IBM it uses somewhere?  - yes. > also under different OS translate, for example: > At https://www.ibm.com/developerworks/rfe/ … R_ID=78176 500 Internal Server Error

69

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pagid, you wrote: P> the Core of applications of a computer at the moment of decision-making on copying IBM-360 is scientific computations. Optimal long words for them it just 48 (either 45, or 46) discharges.  a computer of that time with such long words. Yes, they RISC, somewhat, are not able to process the information byte by byte, and happened the doge have no commands for arithmetical operations with integer numbers. Digit capacity reduction would lead to reduction of high-speed performance by orders. There was such piece, "the mathematical coprocessor". At the time of 8080 it there was separate chip FPU which was engaged in calculations with a floating point. There is no problem to make the same piece and on a computer of the previous generation. Moreover, at a stage of relay/lamp computers it was possible at all  with a floating point and to use analog calculators for which was enough  and a digital-to-analog coder. P> Also what, it could replace at least partly one BESM-6 developed for 20, and can and for 30 years earlier? Most likely yes. Z80 gives 0.5 MIPS, and BESM-6 - 1 MIPS. P> it will cost Only in 1000 times more expensively single-processor, to break time and again in day, and every minute, and hardware resources on join in system it will be necessary more, than in all processors together taken. It under condition of usage of element basis accessible in second half 60. About the price and reliability this your judgement. The easier the processor, the is less in it than logical elements and that it is more reliable. We build in it control of open parity (~10 % from logical elements) and we do the architecture supporting switching-on and switch-off of processors immediately during calculations (modern a cluster it allow). As a result reliability of system as a whole will be above than BESM-like machines which if break, entirely.

70

Re: Re: when the false step has been made? (A sight from 1983)

Hello, , you wrote: > Use. In California And all? About 25 years ago it and in Georgia used. > well here the description of their last hand-made article on 64 discharges (June 2016) on sale, and not just for internal consumption: > https://www.ibm.com/common/ssi/rep_ca/5 … 16-185.PDF z/OC is how I understand, continuation of ruler OS/360 - MVS? Then it is clear. Their OS always included some languages, as that: Algol, Kobol, the Fortran (or even 2), PL/1, . And as mainframes IBM are on sale, it is possible to tell that is on sale PL/1. It is necessary to clarify, whether it is on sale as a separate product. And whether supports IBM it on other platforms.

71

Re: Re: when the false step has been made? (A sight from 1983)

Hello, , you wrote: > there a heap of links of everyones. Except that compiler which in  "Energy", remaining all is mentioned Also the books which have quitted in the USSR, are not mentioned. And the majority of that are mentioned, from 70 - 80th years. One 2009. Then that PL does not develop. Unlike the Fortran and Kobola.

72

Re: Re: when the false step has been made? (A sight from 1983)

Hello, Privalov, you wrote: P> In PL/1 there are pointers. But to name its normal... Links/indexes - necessary, but not "a normality" sufficient condition. PL/1 perfectly illustrates a principle "a camel - a horse created by committee", but anyway if this general programming, instead of bare linear algebra, it was then an optimal choice. If I that did not forget, in the end 60-x there were 2 "normal" languages: Algol 68 and Lisp. Unfortunately the first was too difficult (result of simplification - With and Pascal), and the second too slow.

73

Re: Re: when the false step has been made? (A sight from 1983)

Hello, novitk, you wrote: N> If I that did not forget, in the end 60-x there were 2 "normal" languages: Algol 68 and Lisp. Unfortunately the first was too difficult (result of simplification - With and Pascal), and the second too slow. That that there was in the late sixties has a little the general that now is understood under  which was issued more likely in 80, and  more or less normal received in 90. In the late sixties there were two languages, similar that are now widespread:  - , similar on delfi / early  only without interfaces and the Algol 68 similar on present is fashionable-youth with lambdas (only without parametric polymorphism, unique new that got since then to a mainstream). More precisely not the only thing, without HZ with generations which then too was not.

74

Re: Re: when the false step has been made? (A sight from 1983)

Hello, pestis, you wrote: P> Hello, Privalov, you wrote: whether P>> It is impossible to tear? P> it is possible, why it is impossible. P> 1. Programming languages. With rare exception type of a C and Lisp from the modern point of view it is a pitch gloom. Algol, Kobol, the BASIC, the Hell, Fortran P> 2. Standards. Almost all of them , , are unreasoned and unsuitable for . It is Enough to recall ASN1, teletypes, a G-code and RFC epoch of "paper letters". P> 3. A software. All inconvenient and unreasoned from the point of view of architecture. It is enough to recall RCS, P-CAD, M4, sh, ed and generally unix. It if inside not to look and if to glance there at the best structured programming, and noodles from goto are faster. P> 4. Concepts centralized , mainframes, supercomputers, architecture it is horror flying. P> 5. The approach to project management naive, one only printout of source codes on a paper for archive, without speaking already about  in log. Than a hell is younger, than Si, and is better, unlike more readably. RFC for electronic mail till now it is used. ASN.1 cores, lives, and will live. Unix to you than did not please? Supercomputers and mainframes with their virtual surrounding - as did in IBM, the same approaches are in large quantities used now in data-centers. You , probably.

75

Re: Re: when the false step has been made? (A sight from 1983)

Hello, Klapaucius, you wrote: K> That that there was in the late sixties has a little the general that now is understood under  which was issued more likely in 80, and  more or less normal received in 90. Like Elisp/Scheme came to us from early the seventieth, as well as a pile of other artifacts which have not lived to us of subculture MIT, and are quite normal. K> in the late sixties was two languages, similar that are now widespread:  - ...  added a lot of interesting, but without it it is possible to live till now, therefore I do not rank it as the criterion of "normality". Under "it is possible to live" I understanding development of the big system things of type of OS or languages programming, that is tasks impossible for everyones fortrano-bejsiko-kobolov.