Re: Strange wording about io_t->io_map method

[Marcus Brinkmann]

At Tue, 25 Apr 2006 04:45:37 +0200,
Pierre THIERRY <address@hidden> wrote:
> 
> I found something strange in oi_map description:
> 
>   ``If the invoker has read/write access to the I/O object''
> 
> In capability context, it is rather unseemly. Shouldn't the io_t
> capability have subtypes like read-only, write-only or read/write?
> 
> The text would become:
> 
>   If the capability is a read/write capability, an opaque PATT
>   capability is returned in patt. If the capability is a read-only
>   capability, a weak PATT capability is returned.

Yes, the wording is a bit imprecise.  Thanks for pointing that out, I
fixed it in my local copy.

Permissions are not yet part of the draft, but for I/O objects I
currently envision one permission bit: Write access.  Read access
would always be allowed.

Now that you mention it, for pipes one probably would want write-only
capabilities.  Do you know about other use cases for write-only
capabilities?

Thanks,
Marcus