Hi Oliver,
We often retrieve old revisions of files, but we have never (yet) experienced,
that they were corrupted. I was not aware of about the numerous reports on
"bitrot". That is definately a serious potential problem and in favor of
choosing non-local modes. We are still struggling to get a "real" server setup
with a properly backed up repository, and I will use these reports as input to
management in order to get a real, dedicated CVS server.
Concerning performance: We have setup a CVSNT server on a test machine to check
performance (it was setup as ntserver). I had expected that it would be much
faster. It turned out, that in our case it was not!
Result of test:
checkout all modules: 2m20s using :local:, 2m20s using :ntserver:
update a large binary file: 4s using :local:, 4s using :ntserver:
commit a large binary file: 2m15s using :local:, 1m25s using :ntserver:
That is, only for commit operations the performance was worse (40% longer). I
admit, that these surprising results may be a result of our a network
configuration and the way virus scanning is setup, which is probably far from
optimal.
Concerning your question, what is needed for an application to be "untrusted"
at our place, I admit that this is a stupid classification. An application is
not trusted (at our place), unless it is explicitly stated as being "trusted"
in the specs from the vendor of the file server - and well, cvsnt is not on
that list, therefore it is not trusted.
And yes, if you need the server-side scripts, you should definitely go for a
real server.
The intend of my comment was mostly to tell, that at least in some
configurations, :local: works pretty well.
Regards,
Kim Hansen
-----Oprindelig meddelelse-----
Fra: Oliver Giesen [mailto:ogware@xxxxxxx]
Sendt: 28. januar 2003 15:21
Til: cvsgui@xxxxxxxxxxxxxxx
Emne: RE: [cvsgui] server or local
> In our opinion the DON'T DO IT recommendation is not quite balanced.
>
> We have been using the local mode for more than a year within
> a company with five developers accessing the repository with
> a Win2K file server and Win2K WinCVS 1.3.X clients
>
> We have had very few problems with that.
>
> When it (seldomly) occurs that two users are accessing the
> same RCS file simultaneously, the second client accessing the
> file simply gets a "Waiting for users lock" message until the
> file is released again, wherafter the CVS process proceeds
> with success.
And have you tried to retrieve all old revisions of such files and
verified them? I'm not saying that this happened at your site (and I
never experienced this myself) but there have been quite a few reports
of a phenomenon dubbed "bitrot", i.e. older revisions getting
corrupted over time unnoticed. The possibility of this alone is too
much risk to take IMO.
> For us, the CVS NT server solution is not possible, because
> it is not allowed to install "untrusted" applications on our
> redundant network file server, and installing an NT Server on
> a seperate machine, which accesses the repository on a
> network drive (which is backed up) gives a lot of redundant
> network traffic, which lowers performance.
Then, if you are concerned about network traffic and performance and
especially if you're dealing with files several hundred megs big as
you say, here's some food for thought: Are you aware that neither
zlib-compression nor patching is used at all in local mode? Why should
it? The CVS.exe thinks it is dealing with files on its own file
system.
Honestly, in the beginning I used :local: on a network share myself
for several months (but as the only user). The boost in performance
alone after the upgrade to CVSNT was worth it manyfold IMO. Also,
server-side scripting (which isn't really possible/sensible with the
:local:-share setup unless you distribute the scripts to all clients)
is a very big plus, too.
Just out of interest: What has to happen to classify an application as
"trusted" (enough)?
Cheers,
Oliver
---- ------------------
JID: ogiesen@xxxxxxxxxx
ICQ: 18777742
( http://wwp.icq.com/18777742)
Yahoo! Groups Sponsor
ADVERTISEMENT
<http://rd.yahoo.com/M=244396.2846622.4218523.2848452/D=egroupweb/S=1707281907:HM/A=1414307/R=0/*https://www.clearcredit.com/registration/default.asp?n=b&cpID=c01888p1379&ckID=gen14628>
<http://us.adserver.yahoo.com/l?M=244396.2846622.4218523.2848452/D=egroupmail/S=:HM/A=1414307/rand=464301433>
To unsubscribe from this group, send an email to:
cvsgui-unsubscribe@xxxxxxxxxxx
Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service
<http://docs.yahoo.com/info/terms/> .
[Non-text portions of this message have been removed]
To unsubscribe from this group, send an email to:
cvsgui-unsubscribe@xxxxxxxxxxx
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
|
