Re: [jgit-dev] [egit-dev] Bug 520927: moving away from JSch

["Randall S. Becker" <rsbecker@xxxxxxxxxxxxx>]

Thanks Thomas,

I will look into it when the time comes 😊. Hopefully all will be well, and I won't have to do anything. Otherwise, temporarily grouchy customers and a panicky development group.

Cheers,
Randall

> -----Original Message-----
> From: Thomas Wolf <thomas.wolf@xxxxxxxxxx>
> Sent: November 18, 2018 16:33
> To: Randall S. Becker <rsbecker@xxxxxxxxxxxxx>
> Cc: Dave Musicant <musicant@xxxxxxxxx>; EGit developer discussion <egit-
> dev@xxxxxxxxxxx>; JGit Developers list <jgit-dev@xxxxxxxxxxx>
> Subject: Re: [jgit-dev] [egit-dev] Bug 520927: moving away from JSch
> 
> 
> > On 18 Nov 2018, at 22:05 , Randall S. Becker <rsbecker@xxxxxxxxxxxxx>
> wrote:
> >
> > A key operability requirement, as you mentioned, is not asking for
> > passwords repeatedly. This problem showed up for passphrases when host
> > case sensitivity showed up on ~/.ssh/config. This was fixed. However,
> > the prompting for passphrases should be included. A mechanism for
> > sharing authentication is important for operability within EGit so
> > that users of JGit who add a separate plugin from EGit can avoid
> > reprompting. I suggest that this be made available. It was in JSch and
> > would be a real pain not to have that capability going forward.
> 
> I only understand half of this. The JGit CredentialsProvider is still available,
> and still works in exactly the same ways. The Apache MINA sshd
> implementation is hooked up such that the CredentialsProvider from JGit
> _is_ used.
> 
> A mechanism similar to what happened in JSch _has_ been made available.
> JSch caches keys on the JSch instance; the new Apache MINA sshd
> implementation gives you the possibility to provide a KeyCache to the
> SshdSessionFactory to get similar behavior.
> 
> If you have similar requirements as EGit and want to use some kind of secure
> store, the EGit implementation can serve as a model. AFAIK JSch did not
> include such functionality either. It just avoided multiple password prompts
> because it cached keys under the hood.
> 
> Maybe you just give the new implementation a try?
> 
> Cheers,
> 
>   Thomas