Hi, Paul and All:
Please see my comments that I sent in response to Gosta's comments. There
are some agreements, but there are fundamental dis-agreements the way it has
been used from protocol point of view. It is very dangerous to "hard-wire" a
protocol that views one HLF (or clustered multiple HLFs).
If some one wants to implement that way, it may be left as one of the
choices. But it should not be the only MANDATORY options when we build our
protocol.
It should be rather be ONE of the HLFs (out of many).
I am also enclosing a copy of my earlier email (if someone missed it).
Best regards,
Radhika R. Roy
AT&T
-----Original Message-----
From: Roy, Radhika R, ALARC
Sent: Friday, April 14, 2000 1:27 PM
To: 'Gösta Linder (LME)'; ITU-SG16(a)MAILBAG.INTEL.COM
Subject: RE: [H.323 Mobility:] questions on MTD-016
Hi, Gösta:
I appreciate the clarifications that you have provided: There can be one or
more HLFs in a domain and the H.323 mobile protocol should be transparent to
this. I fully agree with you.
However, I have to draw the attention to your following statement:
"This means that seen from served/visited domains there is just one HLF
entity (clustered or not) to communicate with."
My assumption is that only ONE of the HLFs (out of many because only one of
the HLFs is serving)will be queried to provide the service or answer.
A GK will be able to have an OPTION to choose which HLF to go to. Our H.323
mobility protocol should be flexible enough to provide this OPTION. Please
note that I am using a GK to go to the HLF, not to by-pass it (I also like
to go via GK to VLF, and then via VLF to HLF). Please note that the protocol
should be flexible enough to provide all those OPTIONs.
That is why, I have mentioned in another email that our H.323 mobility
protocol should not be "hard-wired" assuming a single HLF (one or more
clustered together) as if everyone priori knows what that HLF entity is. We
should be careful from protocol point of view.
Let me make it clear by modifying your statement as follows:
"This means from served/visited domains there is only ONE of the HLFs that
will provide the service."
Hope this will further clarify.
Best regards,
Radhika R. Roy
-----Original Message-----
From: Gösta Linder (LME) [mailto:Gosta.Linder@LME.ERICSSON.SE]
Sent: Friday, April 14, 2000 3:02 AM
To: ITU-SG16(a)MAILBAG.INTEL.COM
Subject: Re: [H.323 Mobility:] questions on MTD-016
Hi,
comment on single or multiple HLFs;
To be able to define protocol relates HLF we have to have a common
understanding of the HLF role versus served/visited networks.
In H323 Annex H we have the following related definitions;
- on HLF; "This functional entity is a database in charge of the management
of mobile subscribers. A H.323 Domain may contain one or several HLFs: it
depends on the number of mobile subscribers, on the capacity of the
equipment and on the organisation of the network."
- on VLF; "The VLF contains also the information needed to handle the calls
initiated or received by the H.323 MTs registered in its data base (for some
supplementary services the VLF may have to obtain additional information
from the HLF) the following elements are included: ...."
I would then add that principles for adding HLF entities at the Home admin
domain (for capacity or redundancy or any other reason) should be totally
transparent to the served/visited domains and subject for Home Domain to
reorganise his network according his own needs without reflecting any needs
to reconfigure served/visited domain behaviour nor any data of the external
domains.
This means that seen from served/visited domains there is just one HLF
entity (clustered or not) to communicate with.
I would welcome a discussion to have a common understanding of the
principles to be used according to above, to be able to understand needed
call flows and related messages.
Reagards /Gösta Linder
-----Original Message-----
From: Kumar, Vineet [mailto:vineet.kumar@INTEL.COM]
Sent: den 14 april 2000 01:58
To: ITU-SG16(a)mailbag.cps.intel.com
Subject: Re: [H.323 Mobility:] questions on MTD-016
Radhika,
I have two comments which are embedded in your email below.
vineet
-----Original Message-----
From: Roy, Radhika R, ALARC [mailto:rrroy@ATT.COM]
Sent: Thursday, April 13, 2000 2:03 PM
To: ITU-SG16(a)MAILBAG.INTEL.COM
Subject: Re: [H.323 Mobility:] questions on MTD-016
Hi, Vineet and Steve:
I like to add couple of points with respects to your emails as follows:
HLF:
It is interesting to note that there appears to be an "implicit" assumption
that there is only one HLF in a domain (no matter what happens in the
future).
<vineet:begin> I would agree that there should not be any such assumption. I
don't believe anyone is making such an assumption.
<vineet:end>
Let us NOT make our protocol "hard-wired" like this.
The protocol should be flexible enough to accommodate one or multiple HLFs
in a domain.
If one wants to choose one HLF per domain, it MUST be left as an
implementation issue, NOT a protocol issue.
Home GK:
In the same token, if one wants to declare a GK as its home GK, this option
MUST be provided to a user. It is a choice that MUST be given to a user. If
anyone does NOT like this option, they MAY not use it.
Home Network/Network Address:
In the same token, if anyone wants to declare a network address as its home
network, this option MUST also be provided.
<vineet:begin> Can you please elaborate on the above statement, especially
on "declare a network address as its home
network".
Also, I am not sure about the value of using a term like "home network"
instead of "home administrative domain". The term "home administrative
domain" is already defined in H.323 Annex H and is understood by all
participants ?
<vineet:end>
GK Discovery:
The "old" inefficient GRQ discovery mechanisms may not be the only OPTION to
discover the GK. in a highly mobile cellular environment. An alternative
mechanism, like MGA message can be listened to discover the GK as well.
Best regards,
Radhika R. Roy
> -----Original Message-----
> From: Kumar, Vineet [SMTP:vineet.kumar@INTEL.COM]
> Sent: Thursday, April 13, 2000 4:34 PM
> To: ITU-SG16(a)MAILBAG.INTEL.COM
> Subject: Re: [H.323 Mobility:] questions on MTD-016
>
> Steve,
>
> Thanks for your reply. My comments are embedded in the email below.
>
> vineet
>
> -----Original Message-----
> From: Stephen Terrill [mailto:stephen.terrill@ERICSSON.COM]
> Sent: Thursday, April 13, 2000 4:42 AM
> To: ITU-SG16(a)MAILBAG.INTEL.COM
> Subject: Re: [H.323 Mobility:] questions on MTD-016
>
>
> Hi,
>
> I shall try to answer some of these questions below.
>
> Regards,
>
> ..//steve
>
>
>
> "Kumar, Vineet" wrote:
>
> > Stephen,
> >
> > I have a couple of questions on your contribution MTD-016. These are:
> >
> > 1. H.323 already has mechanisms for discovering the gatekeeper. Are you
> suggesting in your contribution that the terminal should discover the VLF
> instead of the visiting gatekeeper ? Or, are you assuming that the VLF is
> integrated in the visiting gatekeeper ?
>
> This can be discussed - I was of the opionion that we should discover the
> VLF and send the registration to the home environment after that.
> However,
> we haven´t agreed on the role of the VLF, and visited gatekeeper, home
> gatekeerp and HLF - when we come to agree on what these are, my proposal
> may
> change.
>
> <vineet:begin> Yes, this could be discussed in the conference call. My
> suggestion is to reinvent as less as possible. Since we are adding
> Mobility
> to H.323, we should just fill in the gaps that are missing.
> <vineet:end>
> >
> >
> > 2. In H.323, authentication of the terminal and the gatekeeper is done
> at
> the time of discovery. In fact, in H.323 all messages between the terminal
> and the gatekeeper can be authenticated and the message integrity
> preserved.
> In your contribution, authentication is done at the time of registration.
> Why is this preferable to what is already in H.323 ?
>
> I would be interested to understand which gatekeeper you were considering
> should do the authentication. I would assume that the real authentication
> would have to be done at home - as such it would be necessary to find the
> visited network services, and then register/authenticate at home.
>
> <vineet:begin> I agree with your model but my preference is to preserve
> the
> way H.323 works today in that the authentication of both the user (not
> terminal) and the gatekeeper (doesn't matter whether the gk is visisted or
> home) can be done at the time of discovery. The protocol between the
> terminal and the gk does not change regardless of whether it is the home
> gk
> or the visited gk. The terminal sends its hashed password or certificate
> in
> the cryptotoken field of the GRQ message. The visisted gk tunnels the
> entire
> message to the terminal's Home Administrative Domain for authentication
> and
> authorization. There may be an Authentication Function in the Home
> Administrative Domain that authenticates its users. If the user is
> successfully authenticated by the Home Administrative Domain then a new
> temporary password is assigned by the Home Administrative Domain to the
> terminal and the visited gk in GCF for use during that session. This
> exchange is explained in detail in the contribution I made for the last
> conference call. I can't find the MD #. Does anyone know the MD # ?
> <vineet:end>
> >
> >
> > 3. What is the reason for the information flow from the HLF to the home
> gatekeeper, and from the home gatekeeperr to the HLF ? I don't think we
> can
> assume that there is only one home gatekeeper that the terminal may be
> using. In fact, the home gatekeeper may not have any information about the
> user.
>
> I certainly don´t assume that there is only one home gatekeeper. I assume
> that there will be a number of home gatekeepers, but perhaps only one (or
> few) HLFs. In this case, we need an function to select the gatekeeper
> that
> the user is going to camp on - and this may depend on load, subscriber
> profile, policy - or a lot of things.
>
> <vineet:begin> I guess I will have to understand the model that you have
> in
> mind regarding the use of a home gk on which the terminal camps on. In my
> model the HLF will send all pertinent information about its user to the
> VLF.
> The home gk is not needed.
> <vineet:end>
> >
> >
> > Regards,
> > vineet
-----Original Message-----
From: Reddy, Paul K [mailto:paul.k.reddy@INTEL.COM]
Sent: Friday, April 14, 2000 1:31 PM
To: ITU-SG16(a)MAILBAG.INTEL.COM
Subject: Re: [H.323 Mobility:] questions on MTD-016
Hi H.323 Mobility group,
I totally agree with Mr.Gosta Linder's comments on HLF and VLF principles,
that is exactly my understanding.
regards,
Paul
Paul K. Reddy
Intel Corporation
Mailto:paul.k.reddy@intel.com
Office Phone: +1.503.264.9896
Mobile Phone:+1.503.807.9564
-----Original Message-----
From: Gösta Linder (LME) [mailto:Gosta.Linder@LME.ERICSSON.SE]
Sent: Friday, April 14, 2000 12:02 AM
To: ITU-SG16(a)MAILBAG.INTEL.COM
Subject: Re: [H.323 Mobility:] questions on MTD-016
Hi,
comment on single or multiple HLFs;
To be able to define protocol relates HLF we have to have a common
understanding of the HLF role versus served/visited networks.
In H323 Annex H we have the following related definitions;
- on HLF; "This functional entity is a database in charge of the management
of mobile subscribers. A H.323 Domain may contain one or several HLFs: it
depends on the number of mobile subscribers, on the capacity of the
equipment and on the organisation of the network."
- on VLF; "The VLF contains also the information needed to handle the calls
initiated or received by the H.323 MTs registered in its data base (for some
supplementary services the VLF may have to obtain additional information
from the HLF) the following elements are included: ...."
I would then add that principles for adding HLF entities at the Home admin
domain (for capacity or redundancy or any other reason) should be totally
transparent to the served/visited domains and subject for Home Domain to
reorganise his network according his own needs without reflecting any needs
to reconfigure served/visited domain behaviour nor any data of the external
domains.
This means that seen from served/visited domains there is just one HLF
entity (clustered or not) to communicate with.
I would welcome a discussion to have a common understanding of the
principles to be used according to above, to be able to understand needed
call flows and related messages.
Reagards /Gösta Linder
-----Original Message-----
From: Kumar, Vineet [mailto:vineet.kumar@INTEL.COM]
Sent: den 14 april 2000 01:58
To: ITU-SG16(a)mailbag.cps.intel.com
Subject: Re: [H.323 Mobility:] questions on MTD-016
Radhika,
I have two comments which are embedded in your email below.
vineet
-----Original Message-----
From: Roy, Radhika R, ALARC [mailto:rrroy@ATT.COM]
Sent: Thursday, April 13, 2000 2:03 PM
To: ITU-SG16(a)MAILBAG.INTEL.COM
Subject: Re: [H.323 Mobility:] questions on MTD-016
Hi, Vineet and Steve:
I like to add couple of points with respects to your emails as follows:
HLF:
It is interesting to note that there appears to be an "implicit" assumption
that there is only one HLF in a domain (no matter what happens in the
future).
<vineet:begin> I would agree that there should not be any such assumption. I
don't believe anyone is making such an assumption.
<vineet:end>
Let us NOT make our protocol "hard-wired" like this.
The protocol should be flexible enough to accommodate one or multiple HLFs
in a domain.
If one wants to choose one HLF per domain, it MUST be left as an
implementation issue, NOT a protocol issue.
Home GK:
In the same token, if one wants to declare a GK as its home GK, this option
MUST be provided to a user. It is a choice that MUST be given to a user. If
anyone does NOT like this option, they MAY not use it.
Home Network/Network Address:
In the same token, if anyone wants to declare a network address as its home
network, this option MUST also be provided.
<vineet:begin> Can you please elaborate on the above statement, especially
on "declare a network address as its home
network".
Also, I am not sure about the value of using a term like "home network"
instead of "home administrative domain". The term "home administrative
domain" is already defined in H.323 Annex H and is understood by all
participants ?
<vineet:end>
GK Discovery:
The "old" inefficient GRQ discovery mechanisms may not be the only OPTION to
discover the GK. in a highly mobile cellular environment. An alternative
mechanism, like MGA message can be listened to discover the GK as well.
Best regards,
Radhika R. Roy
> -----Original Message-----
> From: Kumar, Vineet [SMTP:vineet.kumar@INTEL.COM]
> Sent: Thursday, April 13, 2000 4:34 PM
> To: ITU-SG16(a)MAILBAG.INTEL.COM
> Subject: Re: [H.323 Mobility:] questions on MTD-016
>
> Steve,
>
> Thanks for your reply. My comments are embedded in the email below.
>
> vineet
>
> -----Original Message-----
> From: Stephen Terrill [mailto:stephen.terrill@ERICSSON.COM]
> Sent: Thursday, April 13, 2000 4:42 AM
> To: ITU-SG16(a)MAILBAG.INTEL.COM
> Subject: Re: [H.323 Mobility:] questions on MTD-016
>
>
> Hi,
>
> I shall try to answer some of these questions below.
>
> Regards,
>
> ..//steve
>
>
>
> "Kumar, Vineet" wrote:
>
> > Stephen,
> >
> > I have a couple of questions on your contribution MTD-016. These are:
> >
> > 1. H.323 already has mechanisms for discovering the gatekeeper. Are you
> suggesting in your contribution that the terminal should discover the VLF
> instead of the visiting gatekeeper ? Or, are you assuming that the VLF is
> integrated in the visiting gatekeeper ?
>
> This can be discussed - I was of the opionion that we should discover the
> VLF and send the registration to the home environment after that.
> However,
> we haven´t agreed on the role of the VLF, and visited gatekeeper, home
> gatekeerp and HLF - when we come to agree on what these are, my proposal
> may
> change.
>
> <vineet:begin> Yes, this could be discussed in the conference call. My
> suggestion is to reinvent as less as possible. Since we are adding
> Mobility
> to H.323, we should just fill in the gaps that are missing.
> <vineet:end>
> >
> >
> > 2. In H.323, authentication of the terminal and the gatekeeper is done
> at
> the time of discovery. In fact, in H.323 all messages between the terminal
> and the gatekeeper can be authenticated and the message integrity
> preserved.
> In your contribution, authentication is done at the time of registration.
> Why is this preferable to what is already in H.323 ?
>
> I would be interested to understand which gatekeeper you were considering
> should do the authentication. I would assume that the real authentication
> would have to be done at home - as such it would be necessary to find the
> visited network services, and then register/authenticate at home.
>
> <vineet:begin> I agree with your model but my preference is to preserve
> the
> way H.323 works today in that the authentication of both the user (not
> terminal) and the gatekeeper (doesn't matter whether the gk is visisted or
> home) can be done at the time of discovery. The protocol between the
> terminal and the gk does not change regardless of whether it is the home
> gk
> or the visited gk. The terminal sends its hashed password or certificate
> in
> the cryptotoken field of the GRQ message. The visisted gk tunnels the
> entire
> message to the terminal's Home Administrative Domain for authentication
> and
> authorization. There may be an Authentication Function in the Home
> Administrative Domain that authenticates its users. If the user is
> successfully authenticated by the Home Administrative Domain then a new
> temporary password is assigned by the Home Administrative Domain to the
> terminal and the visited gk in GCF for use during that session. This
> exchange is explained in detail in the contribution I made for the last
> conference call. I can't find the MD #. Does anyone know the MD # ?
> <vineet:end>
> >
> >
> > 3. What is the reason for the information flow from the HLF to the home
> gatekeeper, and from the home gatekeeperr to the HLF ? I don't think we
> can
> assume that there is only one home gatekeeper that the terminal may be
> using. In fact, the home gatekeeper may not have any information about the
> user.
>
> I certainly don´t assume that there is only one home gatekeeper. I assume
> that there will be a number of home gatekeepers, but perhaps only one (or
> few) HLFs. In this case, we need an function to select the gatekeeper
> that
> the user is going to camp on - and this may depend on load, subscriber
> profile, policy - or a lot of things.
>
> <vineet:begin> I guess I will have to understand the model that you have
> in
> mind regarding the use of a home gk on which the terminal camps on. In my
> model the HLF will send all pertinent information about its user to the
> VLF.
> The home gk is not needed.
> <vineet:end>
> >
> >
> > Regards,
> > vineet
