[comp.unix.questions] SCCS misinformation Rumor-Control

merlyn@iwarp.intel.com (Randal Schwartz) (03/21/90)

In article <2265@tellab5.tellabs.com>, segel@tellab5 (Mike Segel) writes:
|    SCCS uses forward deltas. This means that when you "check in"
|    a piece of code, the original is stored, and the later copies
|    are really deltas based on the original.

Ack. Pthpth. Not this bad piece of misinformation going around *again*.

SCCS uses a method that can get it to any delta *equally well*, (or
equally poorly, depending on your bent).  It doesn't lean in the
backwards or forwards direction *at* *all*.  Just 'cuz the doc sez
something like "noticing the differences" doesn't mean it's storing
them that way.  Internally, it uses something like tagged lines, and
it pulls out all the lines in order that match a particular mix,
computed as a function of the delta that you asked it to generate.

Just another SCCS/RCS/whatever user,
-- 
/=Randal L. Schwartz, Stonehenge Consulting Services (503)777-0095 ==========\
| on contract to Intel's iWarp project, Beaverton, Oregon, USA, Sol III      |
| merlyn@iwarp.intel.com ...!any-MX-mailer-like-uunet!iwarp.intel.com!merlyn |
\=Cute Quote: "Welcome to Portland, Oregon, home of the California Raisins!"=/

runyan@hpcuhc.HP.COM (Mark Runyan) (03/22/90)

>/ guy@auspex.auspex.com (Guy Harris) /  5:41 pm  Mar 20, 1990 /
>> Mike Segel writes:
>>   SCCS uses forward deltas. This means that when you "check in"
>>   a piece of code, the original is stored, and the later copies
>>   are really deltas based on the original.
>
>This is not correct.  SCCS does not use forward deltas.  As Mark Runyan
>stated in his posting:
>
>| I write:
>|  7. RCS stores its revisions so retrieval of the latest revision is quick
>|     and easy, but early revisions take longer.  SCCS stores revisions so
>|     that recovering any given revision takes a constant amount of time which
>|     increases with the number of revisions stored.

I thought I made it clear and I'm glad Guy pointed it out again.  I've
heard many people refer to SCCS's method as using forward deltas, but it
doesn't.  Perhaps if you look at it sideways you could call it forward deltas,
but I think you are misrepresenting it if you do.  It really stores a set
of tagged lines (as another poster pointed out).  The only point being that
RCS is fast on the *latest* revision and slow for past revisions while SCCS
takes a "virtually" constant amount of time for any given revision but that
time increases with the number of revisions.  

Mark Runyan