[sles-beta] php disappeared in beta6?
Joe Doupnik
jrd at netlab1.net
Tue May 20 07:01:55 MDT 2014
Josef,
I have dealt with SLP since it first arose decades ago as a way of
finding printers with duplex printing and colour.
First, do not use multicast or broadcast for services. Those are
dandy on a home/SOHO network. They don't work or survive when routers
are encountered nor when the local wires have lots of other strange
things (think running in the rack of an ISP somewhere). It may work in
one spot but not in the room across the hall (different subnets). Many
servers reject multi/broadcast traffic. Thus dependence on
multi/broadcast is to ensure failure, and you designed it that way.
As stated previously, believing and acting on multi/broadcast
traffic is naive (at best). It is akin to "Click Here." Sound
engineering practice does not countenance such acts.
Second, SLP has the loose notion of SCOPES where both clients and
servers can rendezvous on one of many SCOPE names. That means
configuration details, else failure. It does not scale at all. But you
designed it that way.
Third, SLP requires existing SLP serving software be running on the
network and within radar range of all clients. That becomes an
assumption without which many SLP clients will fail, and you designed it
that way.
DNS has the facilities to resolve names to numbers. It scales. It
is already present. Use unicast to speak to known points.
These matters were understood about twenty years ago. My Brainshare
presentation on SLP is long since lost in the mists of time. The above
points should be sufficient warning to abandon it. Auto discovery is
designed to fail, as I may have mentioned.
Joe D.
On 20/05/2014 12:29, Josef Reidinger wrote:
> On Tue, 20 May 2014 12:18:07 +0100
> "Joe Doupnik" <jrd at netlab1.net> wrote:
>
>> On 20/05/2014 12:05, Josef Reidinger wrote:
>>> On Tue, 20 May 2014 11:57:30 +0100
>>> "Joe Doupnik" <jrd at netlab1.net> wrote:
>>>
>>>> Thanks for chiming in Kai. Your comments simply reinforce
>>>> what I speculated about. Removal has nothing, nada, to do with
>>>> offering more, as anyone can see at first glance. It has
>>>> everything to do with the people who want customers to use the
>>>> "more" facility. Thus the attempts at gloss and justification
>>>> continue.
>>>>
>>>> While here, addressing points by Richard Brown today. Auto
>>>> discovery of services is basically "designed to fail." One would
>>>> not want that rating in one's engineering copybook.
>>>> Broadcast/multicast are normally terminated at routers, often
>>>> terminated upon entry to servers as well. A person/box which
>>>> naively believes what arrives from goodness knows where gets what
>>>> they deserve. The broad/multicast approach went out more than 20
>>>> years ago when MS Windows and Novell IPX traffic flooded our
>>>> networks. May I recommend sound engineering practice here, unicast
>>>> using known destinations. No "Click Here" for our servers.
>>>> Joe D.
>>>>
>>> Well, we use well known
>>> SLP protocol
>>> ( http://en.wikipedia.org/wiki/Service_Location_Protocol ) and it
>>> do multicasting. If SLP (multicasting) is disabled in your network,
>>> then you still can pass on boot command line url to such server.
>> You can't be serious. A server needs booting, and we are
>> supposed to remember to add a particular chunky command line in the
>> middle of that process?
> There is probably misunderstanding. You need it only during
> installation. For massive installation often is used PXE boot with
> predefined parameters. So you need to manually type it only when doing
> installation without PXE, without SLP and without registering against
> scc.suse.com
>
>> Not likely to be accepted in the field. Nor
>> will be SLP; too may variations about its use, and its *cast radar
>> range is severely limited at routers.
>> Long term experience with SLP indicates that it is best ignored
>> entirely. I gave a Brainshare talk on its short comings and offered a
>> better design many many years ago.
> I wasn't on Brainshare and did not hear your talk. Can you share with us
> what is such better design and if is standartized and documented
> similar like SLP with its RFCs?
>
> Josef
> _______________________________________________
> sles-beta mailing list
> sles-beta at lists.suse.com
> http://lists.suse.com/mailman/listinfo/sles-beta
More information about the sles-beta
mailing list