More pedantry
Ah, so the tubes are decatrons.
Web page for the comupter: http://www.scit.wlv.ac.uk/university/scit/history/witch.html
One they are tubes. Mullard would almost certainly have made them (being a UK located company), and they call them tubes. For a set of pictures look here. http://www.tube-tester.com/sites/nixie/datdekat/Z502S_mul/z502s-mullard.htm
Note the label on the box, Mullard made both valves and tubes. If it doesn't control electron flow it isn't a valve. The fact that our American cousins call everything a tube does not mean that we call everything a valve.
Also, it is, rather counter intinuively, perfectly reasonable to express the memory capacity in bits. Everyone gets taught that a bit is a binary-digit. But this is only half the story. A bit is more formally defined as the amount of information that is stored in a binary digit. There are other units of information, three are usually defined: the bit, the nat, and the ban. These are the binary, natural, and decimal units of information. One bit equals ln(2) nats = 0.683 nats, and about 0.301 bans. Thus a decimal digit contains about 3.32 bits of information. Since the computer had 900 decatron tubes each capable of storing a single decimal number (and thus held 900 bans), it contained 900 * 3.32 = 2998 bits of information. The mistake is converting that to bytes. That isn't a well defined operation. A bytes does hold exactly 8 bits of information (and thus also holds 5.45 nats or 2.409 bans), so it is forgivable, but still wrong. Anyway, the conversion yields 373.5 equivalent bytes of information. So something may have wrong anyway. However, not all of the tubes will have been used for storage, many will have been used as computational elements, so it may be that the number of tubes devoted to actual data storage was more in line with the computation. But I suspect someone just got the number wrong.
A little history. The history of the definition of the bit is fully documented. It was first defined and used by Shannon in his seminal paper on information theory: A Mathematical Theory of Communication, The Bell System Technical Journal, Vol. 27, pp. 379–423, 623–656, July, October, 1948. (He gives credit to his college J. W. Tukey for the term.) It is perhaps one of the very few cases where a common term has its entire etymology and history perfectly understood. I get arguments all the time (mostly from IT people, who somehow think that they own the term) that a bit is, and can only be, what a memory element holds. But the guys that invented and defined the term would disagree.
Bletchley Park to restore 112-byte* '50s Brit nuke computer
