[FoRK] binary XML

Mark Day mark_s_day at yahoo.com
Thu Jan 20 07:11:09 PST 2005


I don't think that I have any argument with the idea that there will be some
places in which XML is used without SSL.

In the context of the original question (what fraction of XML-RPC will be
over SSL), I understood your comment as suggesting that there weren't likely
to be many cases: no point in using SSL on a trusted network, better to use
non-SSL on an untrusted network.   That analysis seemed to be quite
different from what I've encountered recently, and I thought it was worth
pointing out the difference.  

These folks running large networks have their reasons for what they're
doing, which may or may not be technically correct.  They are trying to
solve somewhat different problems from the one(s) that you appear to be
working on.

In summary, your mileage may vary.

--Mark

> -----Original Message-----
> From: fork-bounces at xent.com [mailto:fork-bounces at xent.com] On 
> Behalf Of Eugen Leitl
> Sent: Thursday, January 20, 2005 9:59 AM
> To: mday at alum.mit.edu; forkit!
> Subject: Re: [FoRK] binary XML
> 
> On Thu, Jan 20, 2005 at 09:22:14AM -0500, Mark Day wrote:
> 
> > > There's no point to use SSL on a trusted network (within the 
> > > data center) or
> > > over a (static, to avoid setup work) tunnel across 
> untrusted networks.
> > 
> > All consumer-facing transactions of any importance are 
> likely to be over
> 
> Yes, but XML is not limited to consumer. You can use XML over 
> local LAN for
> web front ends and search engines talking to each other. You 
> can have parts
> of a business, or b2b applications which use XML. The 
> consumer only sees
> parts of that infrastructure, if it's not entirely private.
> 
> > SSL.  It's too hard for J. Random User to get VPN tunnels 
> set up right and
> > there's not enough gain over SSL for doing so.  Many B-to-B 
> transactions are
> 
> Not necessarily. When I meant SSL, I meant serving lots of 
> SSL sessions, with
> full PKI session setup. I'd need a real crypto accelerator 
> for that, which 
> costs an order of magnitude more than an entire machine (I'd 
> rather have ten 
> machines in a failover configuration -- I have a dual 
> mini-ITX system for
> some 700 EUR, about 20 W burn total, not counting the disks, if any).
> 
> In contrast, if I ship an on-the-fly prepackaged/generated 
> installer with a shared
> secret, which then sets up an OpenVPN (which is SSL-backed, 
> but gives me full
> access to how I want a VPN set up) session to a cheap (only 
> onboard AES
> accelerator) stack of machines, I can do interesting things 
> on the cheap
> (privacy, VOIP, etc).
> 
> > likely to be over SSL because of the same issue. 
> > 
> > Some large enterprises have decided that they can't really 
> trust their
> > intranet, so their internal applications are gradually 
> moving to SSL.  I
> 
> They seem to not want to distribute secrets via sneakernet 
> (an USB dongle
> with tamper-proof secret onboard and is easily plugged in).
> 
> > know of both a major NY financial institution and a large storage
> > manufacturer who are in this camp. Some hosted enterprise 
> applications only
> > run over SSL -- examples are eRoom and salesforce.com.  
> > 
> > I think all of these folks are likely to at least try 
> deploying some kind of
> > web services, and they'll almost certainly use SSL as the transport.
> 
> Nothing wrong with SSL; but I wonder whether there's any 
> point in running a
> CA cert mint, and how they ship their certs, and how they 
> shield their certs
> (they don't live in the file system, do they?).
> 
> -- 
> Eugen* Leitl <a href="http://leitl.org">leitl</a>
> ______________________________________________________________
> ICBM: 48.07078, 11.61144            http://www.leitl.org
> 8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
> http://moleculardevices.org         http://nanomachines.net
> 




More information about the FoRK mailing list