Pertec interface

Message: 20 Date: Sun, 01 Apr 2012 00:29:09 -0700 From: Jeff Woolsey <jlw at jlw.com> To: cctech at classiccmp.org Subject: Re: Pertec interface Message-ID: <4F7803C5.1050404 at jlw.com> Content-Type: text/plain; charset=ISO-8859-1 SimH also describes bad blocks, which have the high-bit set in the length but otherwise look normal. Except that most tape drives, without heroics, will not give you any data for an unrecoverable read error (both that and a tapemark read as 0-length blocks; some gymnastics are needed to tell them apart), except that such a block cannot be empty. So what to put in it, if we're noting them at all? I'm playing with ANSI-style labels for this. Some history is at: http://neil.franklin.ch/Usenet/alt.folklore.computers/20001209_New_Tape_Con… , though in there it is asserted that tape blocks cannot be longer than 64K, implying that only two bytes are required for the length. However, I have a tape someone else wrote (HDR2 even says so) with 65536-byte blocks (not every O/S today can handle that--16-bit signed comparisons). To represent that along with 0-byte tapemarks requires 17 bits. Elsewhere someone asserted that a tape block can be as long as the entire tape, which seems unlikely and wouldn't always fit in three bytes. I think most Pertec formatted drives will not give you the bad block, but I could be wrong. Of course if it is really bad, even the block length can't be determined. Most PDP-11 and VAX tape controllers, both DEC and third-party could handle 65536 byte blocks. Of course, on a PDP-11, that could be the entire memory in a single tape block! So, it really only made sense on Q-22 systems. Tapes with insanely long blocks are not a good idea, even on hardware that can handle it. I'm guessing 1401 and similar machines could do crazy stuff like that due to their stream of digits architecture. But, having an entire tape with only one LRCC and CRC byte would lead to undetected errors and difficult to recover files. The greatest feature of GCR is that the redundancy data is added periodically within the block, making the likelihood of recovering small dropouts much better. Jon