RE: IPsec Usage Question

[Black_David@emc.com]

> Excerpt of message (sent 31 January 2002) by Black_David@emc.com:
> > In Salt Lake City, I asked folks to become familiar with
> > existing IPsec implementations that they plan to (re)use.  I
> > now have a specific question about this that I need answers
> > to - send the answers to me directly to avoid inadvertently
> > revealing implementation plans (I promise to keep them
> > private).
> > 
> > Q: Does the IPsec implementation you plan to use require
> > 	that the inner IP address be different from the outer
> > 	IP address for traffic that is to pass through IPsec
> > 	to the IP Storage (iSCSI, iFCP, FCIP) system?
> > Follow-up: If so, how do you plan to configure your system
> > 	to securely access a peer IP Storage system from
> > 	another vendor that also has this requirement?
> > 
> > The underlying concern is that requiring that the inner
> > and outer IP addresses always be the same would visibly
> > simplify the configuration required to use IPsec with
> > the IP Storage protocols.
> 
> I'm not sure if this is what you intended, but I'm reading this to say
> that IPsec as used with IP Storage would mandate the same IP addresses
> on inner and outer header.
> 
> If so, that is equivalent to prohibiting external security gateways.
> This is not good.  I understand that there are people who feel that an
> external security gateway is not necessarily the right way to address
> security concerns in IP Storage.  But that's a long way from
> prohibiting their use entirely.
> 
> If that wasn't your intent, could you clarify what you're after?  If
> the goal is to *permit* inner == outer, that's fine.  That's commonly
> supported because that situation occurs when you tunnel from a single
> host to a site protected by an IPsec gateway.  And yes, allowing inner
> == outer in that case indeed makes things slightly easier.

Mandating the same addresses in the inner and outer header is a "big
hammer" that may not be the right course of action.  OTOH, if one
needs to know both the inner and outer IP addresses in order to contact
a target, that has implications for the functionality/usage of Send
Targets, iSNS, and SLP.  My underlying goal is to figure out whether
we need to put support for two IP addresses per target into those
configuration mechanisms (this would apply to FCIP, iFCP, and iSCSI).

I intended to raise the possibility of mandating the same addresses
in the inner and outer header as a possibility as opposed to proposing
it as a course of action as I don't have enough info about what folks
intend to do/think is important to make any proposal in this space.
Paul's input that this would be a bad idea because it would implicitly
ban the use of security gateways is a fine response to that possibility
- and to his credit, Paul was right about the last IPsec-related issue
he brought up (dangling SAs).

Thanks,
--David
---------------------------------------------------
David L. Black, Senior Technologist
EMC Corporation, 42 South St., Hopkinton, MA  01748
+1 (508) 249-6449 *NEW*      FAX: +1 (508) 497-8500
black_david@emc.com         Cell: +1 (978) 394-7754
---------------------------------------------------