root@bu-cs.UUCP (Barry Shein) (07/01/85)
>From: lm@geowhiz.UUCP (Larry McVoy) >Subject: tar.c & blocksize >... >My question is this: is it asking for trouble to play with the NBLOCK >definition in tar.c? Some people here have raised the question of >compatability, ie. what about other sites, what if they don't have source >and/or can't get a hacked version of tar? On the other hand, it seems that >the savings in time are great enough to warrent a change. > >If you have an opinion on this, I'd like to here it. > >-Larry McVoy [ARPA] mcvoy@wisc-rsch.arpa > [UUCP] ...!uwvax!geowhiz!lm As discussed before, there is already a problem. The default blocksize for 4.2bsd tar creates a tape that (as far as I can tell) can not be *physically* read by the tape drive on my SYSV/3B5 due to controller limitations. Making it larger, however, makes it no worse :-) I would suggest you warn your users or, better, not make it the default. You should probably warn your users anyhow, I suggest going no larger than 5 (tar cb 5 ...) if you really don't know who is going to read this tape. As an aside, I still swear I blew the circuit breakers twice on my TU78 by specifying a block size of one at 6250. (Data General also makes it tricky to read tar tapes with physical blocks larger than 8192 though it can be done.) -Barry Shein, Boston University
lm@geowhiz.UUCP (Larry McVoy) (07/05/85)
*** REPLACE THIS LINE WITH YOUR MIND *** I was playing around with tape drives and discovered that a larger blocksize meant faster throughput, so I upped the NBLOCK define in tar.c to 127 (the most blocks our cipher can handle). The results were pretty good, I went from 38 minutes to 18 minutes when dumping about 35 meg. My question is this: is it asking for trouble to play with the NBLOCK definition in tar.c? Some people here have raised the question of compatability, ie. what about other sites, what if they don't have source and/or can't get a hacked version of tar? On the other hand, it seems that the savings in time are great enough to warrent a change. If you have an opinion on this, I'd like to here it. -Larry McVoy [ARPA] mcvoy@wisc-rsch.arpa [UUCP] ...!uwvax!geowhiz!lm