Re: Optimization tips for SSDs ?

[Arnaud Lacombe <lacombar%gmail.com@localhost>]

Hi,

On Sat, Jan 1, 2011 at 11:54 AM, Steven Bellovin 
<smb%cs.columbia.edu@localhost> wrote:
>
> On Dec 31, 2010, at 7:30 39PM, Arnaud Lacombe wrote:
>
>> Hi,
>>
>> On Thu, Nov 11, 2010 at 10:44 AM, Andrew Doran 
>> <ad%homeworld.netbsd.org@localhost> wrote:
>>> git, anyone? :-)
>>>
>> git over ftp/http badly sucks and AFAIK, admins are reluctant to run
>> the git-deamon for resources consumptions' reasons.
>> ftp://ftp.netbsd.org/pub/NetBSD/misc/repositories/git/src/refs/ seems
>> to be badly damaged anyway ... When I tried to use this repo, it did
>> not even kept all the objects integrity, same for the pkgsrc one.
>>
>
> It's not just resource consumption, it's security: why run yet another
> daemon of unknown (and probably unknowable) security properties?
>
That's patronizing a lot of project which use git. At the top of the
list, the Linux kernel, which would certainly not run crap at the
scale they do. I shall also point you to the tentative of corruption
which happened on their CVS[0] export back in 2003. I am more worried
about the relative easiness to introduce backdoor in NetBSD's CVS
repository, which has no checksumming whatsoever than git, which
checksum everything with SHA1. This also prevent data corruption. That
said, git itself can be bugged, as can cvs or gcc, or as, or the CPU
or whatever piece of silicon one is using to automate operations and
the list goes on ...

That said, I do not see the point to continue discussing this, I
already exposed these points here when the tree was created, nobody
cared.

 - Arnaud

[0]: http://slashdot.org/linux/03/11/06/058249.shtml