deprecating msh in favor of hush

Denys Vlasenko vda.linux at googlemail.com
Sat Mar 28 21:28:16 UTC 2009


On Saturday 28 March 2009 21:14, Mike Frysinger wrote:
> ive gone through msh and hush a bit and did a comparison ... here's what msh 
> provides but hush does not in latest trunk:
> 
>  - login
>  - newgrp
> these make me go "wtf are these in the shell?"

It's "design by commitee" for you :) ignore those.
(hmm does bash do them?...)

>  - readonly
> not really important seeing as hush doesnt provide a way to declare read only 
> variables.  afaict, POSIX doesnt require this, so i say we chalk it up as 
> "bashism that'd maybe be nice to have at some point" and see if anyone 
> actually complains.

Yep.

>  - times
> a bashism that looks pretty useless to me and isnt in POSIX.  let it die.

We can implement it as needed.

>  - trap	
> this is about the only thing that hush needs, but we already know that.

Here documents, robust  `cmd`, (cmd;cmd)  and  cmd &  in NOMMU, functions
(because they are easyish to implement when  (cmd;cmd)  is done).

> considering all these things, i say we start the process of deprecating msh.  
> for now, that means simply adding "(deprecated)" to its description in kconfig 
> and telling people to migrate to hush.

Well, just adding "(deprecated)" won't kill anyone. I imagine people
will complain a bit as long as there is anything in msh which
does not work in hush. We can live with that.

> any qualms with this ?  that'd get us down to two shells (ash and hush) and 
> i'm fine with letting those two live.  ash doesnt look like it'd be able to 
> scale down to no-mmu anyways...

Yes, trimming ash down is way harder tha bloati^Wimproving hush.
--
vda


More information about the busybox mailing list