Re: [9fans] patch(1)

Re: [9fans] patch(1)

[ozan s. yigit]

> read the prusker/wobber paper(s).

he means the siphon paper, mid nineties(?) dec labs at europe i think.

oz
---
sense of humor is a vital part of wisdom. -- stanley schmidt

Re: [9fans] patch(1)

[boyd, rounin]

> he means the siphon paper, mid nineties(?) dec labs at europe i think.

yup, the siphon.  PRL [paris] / SRC [palo alto]:

    http://citeseer.ist.psu.edu/prusker91siphon.html

[9fans] patch(1)

[Russ Cox]

On a more positive note, the discussion a few days ago did catch
my attention by pointing out that Nemo had been submitting patches
with patch(1).  I found it so pleasant to apply the patches that I did
the legwork to make patch(1) work for everyone.

So if you have proposed changes to the system, feel free to use
patch(1) to submit them.  It's much more convenient for us, and
I suspect it's more convenient for submitters.

Thanks to David Tolpin for helping to test the setup.

Russ

Re: [9fans] patch(1)

[boyd, rounin]

> On a more positive note, the discussion a few days ago did catch
> my attention by pointing out that Nemo had been submitting patches
> with patch(1).  I found it so pleasant to apply the patches that I did
> the legwork to make patch(1) work for everyone.

patch?  nice idea, but when it goes wrong it goes horribly wrong.

Re: [9fans] patch(1)

[Russ Cox]

boyd, rounin wrote:

>>On a more positive note, the discussion a few days ago did catch
>>my attention by pointing out that Nemo had been submitting patches
>>with patch(1).  I found it so pleasant to apply the patches that I did
>>the legwork to make patch(1) work for everyone.
>>    
>>
>
>patch?  nice idea, but when it goes wrong it goes horribly wrong.
>  
>

just once, it would be nice if you actually understood what was being
said before you started spouting off criticizing it.  see
http://plan9.bell-labs.com/magic/man2html/1/patch,
which has been posted here before.
it's not what you think.

russ

Re: [9fans] patch(1)

[boyd, rounin]

> just once, it would be nice if you actually understood what was being
> said before you started spouting off criticizing it.  see
> http://plan9.bell-labs.com/magic/man2html/1/patch,

i stand corrected (actually i'm seated).

> which has been posted here before.
> it's not what you think.

fix replica/* and i might be amused.

    http://www.cs.bell-labs.com/magic/man2html/1/replica

Re: [9fans] patch(1)

[Russ Cox]

boyd, rounin wrote:

>>just once, it would be nice if you actually understood what was being
>>said before you started spouting off criticizing it.  see
>>http://plan9.bell-labs.com/magic/man2html/1/patch,
>>    
>>
>
>i stand corrected (actually i'm seated).
>
>  
>
>>which has been posted here before.
>>it's not what you think.
>>    
>>
>
>fix replica/* and i might be amused.
>
>    http://www.cs.bell-labs.com/magic/man2html/1/replica
>  
>

what's wrong with them?

Re: [9fans] patch(1)

[boyd, rounin]

> what's wrong with them?

read the archives:

    pull smashed X

read the prusker/wobber paper(s).

Re: [9fans] patch(1)

[Russ Cox]

boyd, rounin wrote:

>>what's wrong with them?
>>    
>>
>
>read the archives:
>
>    pull smashed X
>
>read the prusker/wobber paper(s).
>  
>

i'm not convinced you've actually used replica/*.
you have to explicitly ask it to smash your files.
it's very careful not to smash any local changes unless
you explicitly request that.  much safer than rm, for example.

the problem is that it's too easy to think "well, i haven't made
any changes that i really care about", blindly run "pull -s"
and then all of a sudden your /lib/ndb/local has none of your
useful modifications anymore.  a good (g)ui would help here
considerably.  and maybe overwritten files should get backed up
somewhere.  but the architecture of the system is sound.

siphon is intended for a very different task.  the main difference
is that they're trying to present a single file system, just distributed.
as such the changes propagate in both directions and all the replicas
are assumed to be continuously connected.

replica does software distribution, which is one-way.
usually, i don't want to see the changes that other 9fans make -- if
you edit your /lib/ndb/local, those changes shouldn't go back
to sources.  and there's no way that everyone who downloads
a distribution can be expected to be online 24/7.

russ

Re: [9fans] patch(1)

[boyd, rounin]

> siphon is intended for a very different task.  the main difference
> is that they're trying to present a single file system, just distributed.
> as such the changes propagate in both directions and all the replicas
> are assumed to be continuously connected.

bits of the siphon can do exactly what replica/* does.  i've done it;
managed all of PRL's ULTRIX machines with it.