Re: CVS commit: src/bin/dd

[David Holland <dholland-sourcechanges%netbsd.org@localhost>]

On Tue, Nov 23, 2010 at 08:50:54PM +0200, Antti Kantee wrote:
 > > > > Surely it would be more appropriate to make thye rump kernel directly
 > > > > forward some paths to the real kernel?
 > > > 
 > > > Can you explain how that could work?
 > > 
 > > First thoughts are something like the way /../ is used to 'escape'
 > > from the emulation root.
 > > But maybe just mount the real filesystem within the rump kernel?
 > 
 > The client may be on a different machine, so you'd have to, if not invent
 > a completely new protocol, at least include server side functionality
 > on the client and add some weird pingpong/adjust logic into the server.

But you must already have this protocol in order to be able to reach
the client, no? But if not, we already have nfs, there's no need to
reinvent it.

 > > Making something work well enough to copy files is probably not that hard!
 > 
 > Adding a ton of special case code to copy files is extremely poor design!

Adding weird special case remote access hacks to dd (of all random
tools) is also poor design. Why not for the next round add support for

   dd ifurl=http://www.netbsd.org/index.html of=mycopy

or

   dd ifurl=http://www.netbsd.org/ recurse=true of=mytree/

After all, all you have to do to implement this is open a pipe to ftp
or wget.

 > Doubly so since the only reason is because you don't happen to like
 > some flags which bring 0 cost when not used.  If you don't like them,
 > don't use them.

Poorly thought-out features cause a lot of long-term overhead.

 > Unless someone can actually name a *problem*, move along.

I'd say the first problem is that this wasn't discussed anywhere.

-- 
David A. Holland
dholland%netbsd.org@localhost