Guy Harris wrote:
>
> > No.
> >
> > This was a design decision (look at comments on source code resolv.c) :
>
> Why is that different from IP host name resolution, which uses NIS
> and/or DNS if the OS supports them?
- all modern OS support DNS and/or NIS host/service maps in standard
- hosts are always configured with local hosts/services files or
DNS/NIS support
- this is not the case with ethers maps that are not widely used and
not standard
- the resolv.[ch] code is much easier to maintain if it does not make
assumptions on the underlying OS and so the ethereal behavior will be
the same on all OS (this point is really important for me)
- ethers resolution should be really fast and even faster than host/service
resolution since all packets contain those addresses while IP/ports are
not present in all packets
- since I have implemented ethers file parsing, now it should not be so
difficult to do the same with /etc/hosts and services files parsing and
remove the calls to getXXXbyYYY that could use NIS/DNS, but users will
not be so happy since, IMHO, complete and user-configurable host/services
resolution is much more important than ethers resolution (for IP/UDP/TCP
protocols)
- ...
Laurent.
Once again:
PS: please don't send mail to both me and the dev list. thanks.
--
Laurent DENIEL | E-mail: deniel@xxxxxxxxxxx
Paris, FRANCE | deniel@xxxxxxxxxxxxxxxxxxxxxxxxxxxx
| WWW : http://www.worldnet.fr/~deniel
All above opinions are personal, unless stated otherwise.
