Global Lambda Integrated Facility

Subject Re: Network Description Files for GLIF
From John Graham <johng@xxxxxxxxxxx>
Date Fri, 10 Jun 2005 10:34:38 +0100

Hi Steve - This is a clever idea; and I was thinking how readily extensible it would be to allow for a richer data set and specifically whether different networks and countries might need to express different metadata(?) and how this could accommodated...

I have been attending the TERENA Networking Conference all this week, and as is always the case at these events, the trick is to find the gems that hide in the mass of parallel sessions. I was fortunate enough to happen on a talk by Victorian Giralt (University of Malago, Spain) who was describing how they have implemented a customizable staff directory using LDAP classes and attributes and OpenLDAP Access Control Lists. This allows a staff member to maintain a single set of directory date and to construct rules that determine who can view which columns. So I might want to publicise my cellphone address within my organisation, but hide it for external queriers. I'm no expert in X.500 stuff, but I wonder whether your concept might not be suited to the flexibilty offered by OpenLDAP?

John Dyer, the Chief Technical Officer, was chairing the session and agrees there is some mileage in this. I've copied John on this message in case he isn't part of the GLIF Tech list.

Victoriano's slides can be found here:

I'll be attending iGrid 2005 and possibly also the Internet2 Autumn member meeting in September. In the meantime, hopefully this message will continue the discussions.



Steven S. Wallace wrote:
Here's something I'm working to address similar needs for the Quilt. I think this could scale internationally.

Here's the idea: TXT "us-uk-nl-ca"
FQDN   TXT record showing what country codes contain network map data TXT "IN-OH-FL"
FQDN TXT record showing what states (in the US) have network map data TXT "TFN"
FQDN TXT record showing the name of network(s) in Ohio TXT  "Link=Columbus, OH-Cincinnati, OH" TXT  "Status=Existing" TXT   "Link=Columbus, OH-Chicago, IL" TXT   "Status=Existing"

These are records that specify the network segment endpoints and the segment's status. An application would fetch sequential records until it hit the last one. In this example OARnet (they run the TFN) would be delegated authority for

Each RON would simply maintain their own delegation (zone file). My lab would make an application available that walked the DNS data and constructed a map. Since there may be multiple TXT records for a given FQDN, we could a small set of attributes for each link.

What do you all think?


On Jun 2, 2005, at 12:37 PM, Jeroen van der Ham wrote:

Hello everyone,

In the last two months we started working on a way to make a  description
of our network and resources at Netherlight and Lighthouse, Amsterdam. We
had several goals in mind for making such a description:
- For ourselves, to have a machine readable overview of our network,
possibly including its current (automatically extracted)  configuration.
- For possible users, so that they can see what we resources we have,
possibly what is available at the moment. The description also  contains
the list of possible entry points to our network. Ultimately, these entry
points should point to the network description of the relevant  provider.
- Problem detection, if you have a complete overview of the configuration
of your network, written in a machine-readable format, configuration
problems can be detected more easily. If every network owner were to
publish this in a predefined format, then this can also be used to detect
inter-domain problems, by pointing at the configuration file of the
connecting network in the relevant places.
- Publication in GLIF, to make lambda path setup between participants
easier or even completely automatic.

Yesterday Cees de Laat pointed us at the discussion currently going in
the GLIF community regarding the visualization of the networks.
We think that if everyone in GLIF (automatically) publishes a description
of their network in a predefined machine readable way, this  information
can be gathered by applications more easily. This will make it  easier to
solve problems like the ones Kees Neggers raised in a recent mail
(creating a repository, generating graphics, handling policies).

For example the excellent graphics tool Greg Cole can be used as  is, but
the database is not filled with information entered through a web  form,
but instead automatically by crawlers reading the network  descriptions.
Such crawlers can also be used to periodically collect the network
information and put this in a single repository to provide an  overview.
On the other hand with a machine-readable format, it is also  possible to
parse the relevant information on demand.

We are currently making a draft version of a schema for describing
networks, including information about the capabilities, references to the
polices and services. We will post this to the mailinglist soon.

Bert Andree, Freek Dijkstra, Paola Grosso, Bas van Oudenaarde and  Jeroen
van der Ham.

Dr. John S. Graham
UKLight Technical Manager

University of London Computer Centre
20 Guilford Street

tel:  +44 (0)20 7692 1329
cell: +44 (0)7866 712 817

"Never trust a man in a blue trenchcoat, never drive a car when you're dead."

Tom Waits