GIT vs. CVS vs. SVN

[Kent Borg]

Jerry Feldman wrote:
> Over the years I have used 
> many source control projects ranging from IBM mainframe to SCCS

Ah, the perversion of typing "make" in a completely empty directory and 
having an executable magically appear, as "make" figured out how to 
checkout implicit stuff from SCCS. (Maybe other systems can do that, 
too, but SCCS is where I saw it.)

> I do need to make sure that all the data is 
> backed up properly.
>   

Being distributed, git is actually very good at backing up stuff, at 
least if you have people working together and sharing code. The cool key 
is identifying things by hash, which git guarantees matches the data it 
returns. Say 7fefb92 in my Linux tree and git figures out I mean 
7fefb924d7aed7116fe2a68cdbfc9e36318e7300 (if it had been ambiguous I 
might have to stick on another digit or two), and anyone else with a 
Linux tree knows I am talking about a 27-line whitespace patch to 
pmc551.c, by David Woodhouse, in September of 2006. So though I can rain 
great confusion on my own tree, if I still have the hash tag and you 
still have the hash tag, I at least know my copy of the files matches 
your copy at that hash. And I know the hash of the ancestor and that it 
will match your hash of the ancestor. (And so I can work all the way 
back to 1da177e4 with certainty that, if I have the data, I have the 
right data.) Combine that with it being so easy to type "git clone" and 
"git push" and your data can be pretty durable.


-kb, the Kent who is still learning git (and who still uses rcs, on /etc 
files).