Re: [Gnu-arch-users] [MERGE-REQUEST] Support for several SSH implementations

[Ludovic Courtès]

"Alfred M\. Szmidt" <address@hidden> writes:

> One small comment, make the variable names a bit more generic and
> descriptive and preferbly use an existing name.  For example,
> TLA_SSH_BINARY is bad since another GNU Arch implementation could
> implement such a feature.

I wondered, too.  Probably `ARCH_' would be a better prefix.

> Adding _BINARY is redundant too.

No, it's not.  By default, the program name is determined based on the
SSH implementation type.  I.e., the default program name for
implementation type `lsh' is `lshc', the program name for `openssh' is
`ssh', etc.

But sometimes you'll want to specify another program name, like `lshg'
for implementation type `lsh', or, say, `/usr/openstuff/ssh' for
`openssh', hence the `_BINARY' option.

> Since you could in theory end up using plain old remote shell, or a
> variant of it that encrypts the traffic or anything inbetween.  I'm
> also skeptical over the RSH_TYPE thing too, cannot this be specified
> in some other way? By only using the RSH variable?  Not even sure what
> that variable does...

That's what I thought at first but it is not true in the current
implementation: tla really expects some implementation of SFTP, not an
arbitrary RSH implementation.

Theoretically, I agree that a generic RSH layer would have been, well,
more generic.

> Wonderful patch though, I prefer using lsh over ssh so I'll enjoy this
> one. :-)

So do I.  But note that (on PPC at least) an `abrowse' with `lshc' or
even `lshg' turns out to be much, much slower (around 12 times as slow
as OpenSSH!).

Thanks,
Ludovic.