Skip to Content.
Sympa Menu

per-entity - RE: [Per-Entity] HTTPS transport and TLS trust

Subject: Per-Entity Metadata Working Group

List archive

RE: [Per-Entity] HTTPS transport and TLS trust


Chronological Thread 
  • From: "Cantor, Scott" <>
  • To: Scott Koranda <>, "" <>
  • Subject: RE: [Per-Entity] HTTPS transport and TLS trust
  • Date: Tue, 6 Sep 2016 14:06:40 +0000
  • Accept-language: en-US
  • Authentication-results: spf=pass (sender IP is 164.107.81.222) smtp.mailfrom=osu.edu; gmail.com; dkim=none (message not signed) header.d=none;gmail.com; dmarc=bestguesspass action=none header.from=osu.edu;
  • Ironport-phdr: 9a23:ypjhKhOirqvLIlWzgrAl6mtUPXoX/o7sNwtQ0KIMzox0KPn7rarrMEGX3/hxlliBBdydsK0UzbeN+Pm9EUU7or+/81k6OKRWUBEEjchE1ycBO+WiTXPBEfjxciYhF95DXlI2t1uyMExSBdqsLwaK+i76xXcoFx7+LQt4IPjuUs6X1pzvlrP6x5qGWA5PmCb1QrRoJRO6pE2FrcoRm4JkJq8ZxR7Ao38OcONTkycgb0qehVPx4Nux4I9L8iJbvPck8MgGVr/1NexsVbFCSTkgL2Ep9eXqsxLESA6I4D0bSGpAwTRSBA2QpjrzWIv2qG+yjeF6xDXQdZn9RLYoXi7ktY9sUwKuhSsaYW1quFrLg9B92foI6CmqoAZylsuNONmY
  • Spamdiagnosticmetadata: NSPM
  • Spamdiagnosticoutput: 1:99

> Are there other arguments in favor?

The generic one is just to limit the chance of a stale result being injected
into the stream. Without real time signing, you don't have much of a
freshness guarantee, so there are still windows during which stale metadata
responses could be injected. XML Signature doesn't natively have properties
related to replay or freshness. Using TLS raises the bar for that attack, but
less so if we're talking CDNs since that basically eliminates any real hope
of verifying the TLS certificate at any strength.

> What other objections are there to serving InCommon metadata
> using the TLS trust model?

Mainly the lack of trust constraints. If we assume that ADFS is validating
the certificate correctly and performing revocation checking, it still
validates any certificate it gets against any valid trust root, which means
there's no control over which trust anchor is trusted for specific sources of
metadata.

Even if you do the self-signed thing, you still have to trust that
self-signed cert for basically anything in order to trust it for that
specific case.

The basic objection is over whether people would leverage the TLS layer
correctly or effectively and whether that's a thing to even encourage given
the complexity of getting it right.

-- Scott




Archive powered by MHonArc 2.6.19.

Top of Page