Talk:Comparison of early computing machines

Colossus: binary?

Nice table, but I think we can assert that Colossus was "binary". It used decimal counters, but most of the processing occurred on binary streams 5 bits wide. An operator could wire up AND, OR and XOR functions in any combination. — Matt Crypto 09:09, 26 August 2006 (UTC)

Digital column

Is this column even necessary? The title of the table already indicates that all these computers are digital. I don't think it's necessary to reiterate that point. -- mattb @ 2006-11-11T06:15Z

I went ahead and removed it per my reasoning above. -- mattb @ 2006-11-11T19:48Z

Nation

What's the general view about the need for a "nation" column? It isn't really a "defining characteristic", as the table's heading suggests it should be. I also feel it may detract a little from the impression of an article that aims to present an impartial account of the development of computers. It's interesting, of course, from the point of view of nationalistic sentiment. But people can find out where the computers were developed by clicking on the links, without the need to emphasize this by putting it in a table about computing characteristics of the machines. What does anyone else think? Adrian Robson 22:32, 8 December 2006 (UTC)

I agree. I don't think there's any particularly good reason to leave this information in the table. -- mattb @ 2006-12-14T20:14Z

Thanks. As there have been no other comments over the past month, I've removed the nation column. Adrian Robson 08:57, 12 January 2007 (UTC)

It seems to have reappeared with no discussion! Pterre (talk) 14:18, 9 February 2008 (UTC)

Should Stibitz's Complex Number Calculator be added?

In the article on Stibitz it says he was "the father of the modern digital computer". I think that sentence should be modified anyway, but it still might be worthwhile to add the "Model K" or the "Complex Number Calculator" to the template. Christian Storm 14:22, 8 June 2007 (UTC)

Reverting to very old version – restoring POV

This template has outgrown its usefulness. Neutrality-pushers are removing any pedagogic usefulness (colors) form the table as POV. At the same time it seems that every nationalist wants to include their countries contribution to the table. The table now contains 11 “early” computer instead of the original five!

There is a point-of-view, shared now by everyone involved with computing, that a useful digital computer must be electronic, binary, programmable and Turing complete. The original intent of this template was to show that none of the five early computers, in their six incarnations before 1949, where all of these. Everyone of them failed miserably on at least one point. The IBM ASCC, the direct ancestor of todays IBM line of computers failed on 3½ of 4.

At the same time, all four of the requirements were met by at least one or two of the early computers. While none of the computer can claim to be the father of modern computers, they can all state equal claim to being one of the grandfathers. In fact, the five computers were listed, not because they were the first, but because they all contributed something important to the development of modern computers.

The last useful version of the template is from 25 October 2007 by Matt Crypto.

Computer	Shown working	Place	Binary	Electronic	Programmable	Turing complete
Zuse Z3	May 1941	Germany	Yes	No	By punched film stock	Yes (1998)
Atanasoff–Berry Computer	Summer 1941	USA	Yes	Yes	No	No
Colossus	December 1943 / January 1944	UK	Yes	Yes	Partially, by rewiring	No
Harvard Mark I – IBM ASCC	1944	USA	No	No	By punched paper tape	Yes (1998)
ENIAC	1944	USA	No	Yes	Partially, by rewiring	Yes
	1948	USA	No	Yes	By Function Table ROM	Yes

In the very next edit neutrality-pushers start eroding the useful POV of the template by removing colors. In the history of computing decimal is not of equal value to binary!

-- Petri Krohn (talk) 23:43, 27 May 2009 (UTC)

P.S. - This is not an article, it does not need to be NPOV. It is a table used to illustrate a certain point-of-view. -- Petri Krohn (talk) 11:08, 28 May 2009 (UTC)

That pov needs to be clarified, as there now seem to be at least a couple of glaring omissions in the table. --Malleus Fatuorum 20:30, 28 May 2009 (UTC)

Outgrown usefulness?

Has this template really outgrown its usefulness? The table has evolved to give a historical context within the articles about most of the machines that it includes, as well as the 'parent' article the history of computing hardware.

The NPOV principle is surely fundamental to Wikipedia and should apply to templates that are incorporated into articles as well as to the articles themselves. Petri Krohn argues that this table is used to illustrate a certain point of view. It is not clear what this POV is—is it that green indicates features that carry forward to today's computers and red indicates the 'dead end' features? If so, it should be more explicit, and why omit from the table the first computer(s) that score all green cells in the table (Manchester Small-Scale Experimental Machine, EDSAC, Manchester Mark 1 and CSIRAC?

Petri Krohn states that 'electronic', 'binary', 'programmable' and 'Turing complete' are the four defining characteristics of a useful digital computer. I would put 'stored program (in a read-write memory)' ahead of 'Turing complete' in such a list.

We need a substantial debate before making a reversion to a version that is 20 months old.

--TedColes (talk) 14:20, 28 May 2009 (UTC)

• While I wholeheartedly agree with pruning this table, like TedColes I'm concerned about the apparently arbitrary exclusion of the SSEM, the first machine that recognisably ticks all of the boxes expected of a modern computer. The Manchester Mark 1 in addition ticks a box that probably ought to be included, two-level storage (RAM and magnetic backing store), again an important milestone. --Malleus Fatuorum 20:25, 28 May 2009 (UTC)

In defining a pruned-down version of the table, it is difficult to define a 'natural' boundary around its contents. I chose to try to be inclusive for the decade of the 1940s in the set of machines included. Another boundary might be 'program-controlled', which would exclude the George Stibitz machines and the ABC, but there would not be a natural finishing point, and doubtless different views as to where the list of machines should end. There is another concern about defining an appropriate set of criteria for the columns, and whether or not to use colours. Perhaps we should start with agreement as to the purpose of the table.--TedColes (talk) 09:37, 31 May 2009 (UTC)

Delete 'Turing Complete'?

I question whether the concept of 'Turing Completeness' is appropriate for this template and the articles into which it is transcluded. In the strictest sense, no finite machine is truly Turing complete.

Professor Benjamin Wells of the Departments of Computer Science and Mathematics, University of San Francisco, has recently published a peer-reviewed paper which shows that a Universal Turing Machine could have been run on the set of Colossus computers. This means that Colossus satisfies the definition of 'Turing Complete' given in the Wikipedia article Turing completeness.

Wells also points out in the paper's Footnote 7 that a Universal Turing Machine is not the same as a 'general purpose programmable computer' and Colossus was clearly not that.

What do others think?

--TedColes (talk) 16:19, 10 November 2009 (UTC)

Should Z4 be included?

TedColes deleted my entry regarding the Z4 in the Template:Early computer characteristics. Enquire (talk) 08:10, 12 December 2009 (UTC)

According to the website of Horst Zuse (Konrad's son) [1] "The Z4, the second general purpose computer, was almost completed in 1945. The Z4 was an customer order by the Henschel Aircraft company in 1942. However, it was not possible for Konrad Zuse to get the Z4 operative before his escape from Berlin to Hinterstein March 16, 1945." and "From July 11, 1950, this configuration was used for five years at the Institute --TedColes (talk) 17:44, 22 June 2010 (UTC)of Applied Mathematics at ETH Zurich."

I have therefore removed this entry from the table of computers that were operational in the 1940s.--TedColes (talk) 10:05, 11 December 2009 (UTC)

Whether or not it was fully working in 1944 is an interesting point, however, this was designed in the early 1940s and was reassembled in 1945 and later delivered in 1950. I think one can assume that it was working well before it was delivered. As such, I do not see this as grounds to eliminate the entry. Rather, a comment about the delay in delivering the unit, which was obviously delayed by WWII and nothing to do with the technical merits or limitations of the technology. I therefore feel strongly that it should be included, with the relevant notes, and not obliterated from the record as you propose. Enquire (talk) 22:07, 11 December 2009 (UTC)

Also, this template name is {Template:Early_computer_characteristics} and, although it has a subtitle "Defining characteristics of some early digital computers of the 1940s" I have not found similar templates for the 1930's and 1950s ... and in context, this template is presented as a history of the early computers. Therefore to eliminate the Z4 and, for that matter, the Z1 and Z2, possibly even the Z5 is a misleading and incomplete chronology of the birth of computers (which is how most articles present this template). Something to think about. I am not going to get into an edit war over this, but I do feel that this needs to reviewed on the broader context of what was happening in the early part of the last century in terms of how the computer was conceived and evolved into what we have today. To pick and choose which bits of that chronology we include and which bits we toss out is rather arbitrary and defeats to objective of being a complete record. Researchers look to Wikipedia for a comprehensive understanding and may not appreciate that some things were removed for reasons that to them may seem arbitrary. Put the comments about the removal and rebuilding of the machine, but don't delete it entirely. It is a 1940's computer and it was almost certainly working as such during the latter 1940's ... even if it was not delivered to the client until 1950, that is a contract issue, not a technical one. Enquire (talk) 22:07, 11 December 2009 (UTC)

If Wikipedia is to be a resource of all knowledge, I do not think it fits the mandate to entirely delete a significant development on a pure technicality. In any event, computer design did not begin, nor did it end in the 1940s, although that was a significant decade. Besides, the Z4 was designed and built during the 1940's and there is no reason to believe it was not fully functional by the late 1940's, even though it was delivered to the customer until 1950 ... which as I wrote earlier is really a contractual issue, not a technical nor substantive issue from the perspective of the history of computers. Enquire (talk) 08:10, 12 December 2009 (UTC)

Enquire raises a question here that relates to the purpose of this table. I am coming to the view that, as it clearly cannot be all things to all editors, it needs a good definition of its purpose. I would argue that this purpose should be to provide a broad context of computing developments in the 1940s, but it should exclude developments that did not become operational before 1950. The context would therefore be useful in the articles into which it is transcluded. By 'become operational' I mean that the system was performing useful work for the end users, not merely for the developers. The column headed 'First operational' similarly is the date when that came about.

Furthermore, I think that there is a case for excluding devices that were not programmable, and also, as I suggested above, to remove the 'Turing complete' column as Turing completeness does not equate with 'general purpose'.--TedColes (talk) 18:32, 12 December 2009 (UTC)

We now have some more definitive information about the Z4 and when it became operational. This editor put the following on my talk page, but I think it is worth displaying here too. "The machine Z4 was operational in March 1945 in Göttingen, and, after being stored in Hinterstein from 29th of APril 1945 till the end of 1946 in Hinterstein, it was operational again in the beginning of 1947 in Hopferau in the Allgäu." (my translation) (in German: "Die Maschine Z4 war im März 1945 in Göttingen funktionsfähig und nach einer Einlagerung vom 29. April 1945 bis Ende 1946 in Hinterstein war die Z4 Anfang 1947 funktionsfähig in Hopferau im Allgäu.) So there slipped an error on Horst Zuse's English page. F.L. Bauer also writes, that it was transported to Göttingen in Feb. 1945. 217.236.214.106 (talk) 17:03, 22 June 2010 (UTC) --TedColes (talk) 17:44, 22 June 2010 (UTC)

Scheutz's calculator

If the Atanasoff–Berry Computer is included shouldn't the Scheutz's calculator also be included? And by the way I think the Turing complete column should be removed, practically anything can be made Turing complete. The Z4, modified ENIAC, and some of the machines at the end of the table could be removed if space is at a premium. I think though the Z1 should be included as the only reason for its removal is that it was so unreliable it only worked for a few minutes at a time. The very high unreliability should of course be noted though Dmcq (talk) 17:20, 22 June 2010 (UTC)

Neither Scheutz's machine nor the Z1 was a 1940s machine. It is unfortunate that a US judge said that the Atanasoff–Berry Computer was correctly named as such, as many would regard it to be a calculator by today's standards.--TedColes (talk) 21:48, 22 June 2010 (UTC)

The title of the template is 'Early computer characteristics', the heading of the table is easily corrected to before 1950 or some earlier date. Dmcq (talk) 21:53, 22 June 2010 (UTC)

Why do you want to remove the Turing-complete column? It's the very characteristic which makes a machine a universally applicable computer. The first universally complete formalism was not the Turing machine, but Gödel's universal formalism in his completeness proof in 1930, and a further one in his incompleteness theorems in 1931.

Babbage's machine would have been Turing complete but could not be built at his time.

So the first actual realization of Turing completeness is historically interesting. And by the way -- the Z3 was proven to be programmable in such a way that it is Turing complete in 1998 by Rojas ... that is no problem, that the proof was belated -- the first proofs of the universality of actually built computers (quote) "seems to have been first written down by Hermes, who showed in (...) how an idealized computer could be programmed to duplicate the behavior of any Turing machine" [source: Shepherdson and Sturgis].

This was a paper by Hans Hermes (in 1954, 8 years after the ENIAC was operational).

And why do you want to remove the Z4? It was an extended version of the Z3 which had branching (cf. Rojas), it survived the war, it was the first or second commercial computer in the world, it is the first Turing complete machine to survive to this day, it inspired the first Swiss computer (at the ETH) -- that should suffice concerning its importance.

Sincerely, 217.236.206.77 (talk) 10:43, 23 June 2010 (UTC)

PS It would actually make sense to write the year of the proof behind the yes in the column -- yes (1954). 217.236.206.77 (talk) 10:52, 23 June 2010 (UTC)

PPS When Zuse was working on his computers, he was also working on his Inception of a universal theory of computation with special consideration of the propositional calculus and its application to relay circuits ("Ansätze einer Theorie des allgemeinen Rechnens unter besonderer Berücksichtigung des Aussagenkalküls und dessen Anwendung auf Relaisschaltungen" (1943), unpublished manuscript, Zuse Papers 045/018) so it's easy to see that he had universality in mind. 217.236.206.77 (talk) 11:03, 23 June 2010 (UTC)

The programmability and it being a complex computational task and that it worked digitally rather than being an analogue computer is what's important. Turing completeness is unimportant and not all that interesting and too easy to achieve with tricks, I'm pretty certain Colossus could be proved Turing complete for instance but that isn't very interesting. If others feel something for it I'm not going to dispute them though. The bit I really don't see as reasonable is having a start point of 1940. And by the way it is pretty clear to me that Babbage's machine could have been built and made to work with the materials and techniques of the time, the real point is that it wasn't actually built. The Scheutz's calculator and some of the more complex automatic tabulators should qualify if the Atansoff-Berry computer is included. By the way the earlier Zuse machines had branching in their microcode sequencers. The later machines just take up room and don't add significantly to the information - people can always look up the appropriate article as noted at the top. Dmcq (talk) 11:26, 23 June 2010 (UTC)

For a formalism or a machine to be μ-recursive / Turing complete or not is without any doubt a fundamental difference -- just look at the history of theoretical computer science.

And as my last comment showed, the way to program a machine in such a way that it can simulate a Turing machine was proved later anyway -- for the ENIAC, too. But that doesn't change the fundamental difference. How sure are you, that you could use a "trick" to make the other machines Turing complete, or is it just guessing?

We certainly don't want to include mechanical calculators -- like Leibniz' Stepped reckoner - (which worked digitally), or punch card controlled machines like the Jacquard loom (programmability), do we? Not only programmability or working digitally is important, but the internal computer organization that moves the machine closer to being a universal computer.

So the year 1940 is not arbitrary, because of the Z3 in 1941. (if one demands branching – the Z4 in 1945) The decade which the table covers right now is really the very decade that saw the first universal computers; it makes perfect sense.

If you disagree, could you please be more specific for what reasons you want to include certain machines and for what reasons you deem certain characteristics to be important and others not?

Sincerely,217.236.206.77 (talk) 15:08, 23 June 2010 (UTC)

PS One could add to the Turing complete column the supplementing info, that the SSEM was the first computer which was universal without the need for laborious program transformations – this is something which Raul Rojas wrote (who proved the Turing completeness of the Z3). 217.236.206.77 (talk) 15:17, 23 June 2010 (UTC)

(See "Delete 'Turing Complete'?" above which cites a paper that shows that Colossus was indeed Turing complete.)--TedColes (talk) 15:27, 23 June 2010 (UTC)

Interesting. This makes a -- yes (2009)

I think the way to deal with this issue is to include appropriate information, maybe in footnotes; so to inform the reader - when was it proven? How practical was its universality? Etc. Like Rojas wrote - that the SSEM was the first computer that was practically universal. 217.236.206.77 (talk) 16:06, 23 June 2010 (UTC)

So someone has done the Colossus. It is still not interesting, branching for instance was something that Zuse was aware of and implemented where it mattered, but it didn't matter or could be worked round easily in floating point computational loops. As to programmability and decision making automatic tabulators did that before Zuse even see [2] for an example. The only thing they left out was going in a loop, sticking the output cards into the input I'm pretty certain one could probably quite easily produce a Turing machine with them. Practical universality sounds a better condition as a stage towards modern computers but I can see arguments about its definition. The reason for excluding Z1 that it was a prototype and wasn't reliable enough to be of practical use just isn't reasonable that I can see. Dmcq (talk) 17:21, 23 June 2010 (UTC)

Of course some kind of programmability was implemented before -- that’s what I wrote -- the Jacquard loom used punch cards already.

Hermes proved the Turing completeness for the early computers in 1954, Rojas for the Z3 in 1998, Wells for the Colossus in 2009 ... now "you're pretty certain" - well, that doesn't really convince me.

It's been a while since I worked through this theoretical kind of stuff, but I'm sure I could follow such a proof. If it's so trivial to make the other machines Turing complete -- go for it.

... Ok, you deem it not interesting. Others do. And Turing completeness is a mathematically sound concept; that's why I've no doubt that it makes sense to include it.

Then in a footnote a sentence about the practical universality can be added.

Sincerely, 217.236.206.77 (talk) 17:40, 23 June 2010 (UTC)

PS According to all I've read, the Z3 was not simply a semantically equal replica of the Z1. Zuse refined the internal organization of his computers. So Rojas did not prove that the Z1 was Turing complete. Thus it should not be in the table. 217.236.206.77 (talk) 17:43, 23 June 2010 (UTC)

Hero of Alexandria made a programmable machine nearly two thousand years ago, so it's nothing new, I wasn't saying it was. If you're really set on the Turing machine business then I suppose it doesn't take up too much room. The main change in the functionality of the Z3 compared to the Z1 is that it handled infinities and NaNs in its floating point otherwise it was very similar logically. Anyway I see no argument excluding the Z1, automatic tabulators or the Scheutz calculator that wouldn't remove some machines already in the table so either the table should be trimmed or some earlier machines added or preferably both. Dmcq (talk) 22:43, 23 June 2010 (UTC)

So, we are back to the old question of what distinguishes a computer from a calculator. Is there a satisfactory answer to it?--TedColes (talk) 06:37, 24 June 2010 (UTC)

Well, if we want to start differentiating between different machines and languages we can use mathematical concepts of computability and state the logical equal of different formalisms, of formalisms and machines ...– once they're μ-recursive one kind of reached the top. This at least is a sound characteristic.

The idea of practicability or practical universality is much harder to agree upon; I think in this respect one of the best results we can get to is gathering some quotes, and put them in the footnote, like e.g. Rojas, who, when examining exactly these questions, came to the conclusion, that the Manchester SSEM was the first practical universal computer.

Because since it is not possible to actually prove this kind of subjective quality, IMHO the best we can do is gather some experts' assessments; and make sure for the reader : - there simply is no absolute agreement. Opinions differ.

That's also the very reason why I think it's not bad to include so many machines in the table -– the creation of the computer simply was no single event, but an evolution.

217.236.175.147 (talk) 15:10, 24 June 2010 (UTC)

I'd have said the important point was that it could be programmed to automatically do a substantial calculation. And to satisfy the digital crowd that it was digital rather than analogue though I don't see what people have against analogue computers. That a computer was acknowledged to be a practical Turing machine could be used as an end point.

I was wondering what the template was used for and the places the template is used are:

Analytical engine, Atanasoff–Berry Computer, Colossus computer, Computer, CSIRAC, Electronic Delay Storage Automatic Calculator, ENIAC, History of computing hardware, Harvard Mark I, Z3, Z4

The template misrepresents the state after the Analytical machine and something more earlier would be better for the history of computing hardware and computer. For the rest it is more of a brag box and more earlier would give better context.

The template is about 'early' computer characteristics. So it should include those in a more historical order of new characteristics if anything.

As to analogue machines you can have a look yourself and ask why they are being excluded for Tide-predicting machine and Differential analyser. By the way you can see some Meccano implementations of Babbages machines at [3] including the Difference Engine 2 and the control mechanism of the Analytical engine, plus there is a quite large Differential analyzer and a good bibliography. Dmcq (talk) 16:45, 24 June 2010 (UTC)

BINAC operational?

My understanding is that BINAC never became operational in the sense of serving the end users, rather than the developers. Is there evidence to the contrary? --TedColes (talk) 20:51, 14 September 2011 (UTC)

I believe you are correct in that BINAC was never used operational however it was run successfully at it's manufacturers. On one occasion running for several hrs without problem. I think it should be included in the list for two reasons :

1)The ABC and Manchester baby were never operational, The ABC never solved it's storage problem and the Manchester Baby was a prof of concept and only used to run a few programs. They are both in the list 2)The BINAC was the second stored programme computer to run and the first in the United States. It was also the first commercial computer. — Preceding unsigned comment added by Rknigh21 (talk • contribs) 16:50, 16 September 2011 (UTC)

Suggestons

I suggest this new version, which was reverted due to too extensive changes. (A column about Von Neumann architecture was added. More first generation computers (1940s and early 1950s) were added, especially the first von Neumann architecutre machines, and also the first with magnetic core memory and drum memory, and the most famous Scandinavian and Russian machines. I have embedded the template in those computer model articles, but since they were removed from the article, now the template inclusion looks kind of strange.) Mange01 (talk) 23:56, 12 January 2012 (UTC)

Inexact information? Computer Historian needed

I have copied the following from the ENIAC talk page: --TedColes (talk) 12:10, 14 February 2012 (UTC)

The comparative table of the first computers contains some information that may be inexact.

1. ENIAC became operational in November 1945, when it ran a series of calculations for the hydrogen bomb. ENIAC was secretly operating a few months before it was introduced to the public in July 1946.

2. According to the book "Origins of Cyberspace: A Library on the History of Computing, Networking and Telecommunications" by Diana H. Hook and Jeremy M. Norman, Zuse Z4 did not have conditional branching in its original 1945 version. Z4 became Turing-complete only in its final version, which became operational in 1949-1950. Well, one may argue that the same odd hack that is described in Raul Rojas's paper on Z3's purely theoretical "Turing-completeness" could work also with Z4, but that's not the way how Konrad Zuse himself had implemented the branching later.

3. I am not sure that "Modified ENIAC" deserves a separate entry in the table. All those computers, including ENIAC, were modified significantly a number of times through their history, but there was no such machine called "Modified ENIAC". These modifications were just a part of ENIAC's development, followed by a series of other modifications.

I am not going to change any of these details, but I think someone who is an expert in computer history should check it out. Laplandian (talk) 20:48, 14 December 2011 (UTC)

I agree that it would be appropriate for someone who is an expert on the history of computing to improve the table. --TedColes (talk) 12:07, 14 February 2012 (UTC)

Deletion proposal

This template is being considered for deletion - see [Wikipedia:Templates for discussion] I disagree but think that it might benefit from being slimmed down. I would urge those with views on this to comment at the entry on the Templates for deletion page. I suggest that the slimming down might include reducing the entries for Colossus and ENIAC to single rows and removing the column for 'Turing Complete' - an obstruse matter for most readers and not the same as 'General Purpose'.--TedColes (talk) 07:48, 20 April 2012 (UTC)