I'm not sure we can ever flag an untrusted authority as an error as a rule 
because a tenant could be creating a DS explicitly with internal trust which 
could be fine if their application stack could consume it.  It's basically just 
informational to know if the cert is in a global trust or not.  We could return 
a 200 with a warning maybe.  A lot of our endpoints use following response 
idiom:

"alerts": [
    {
      "level": "success",
      "text": "Profile was deleted."
    }
  ]

Seems like today we basically just map 200 == level success and basically 
everything else == level error.  What would the TP UI do with that information 
is a different question.

I'd be ok with just dropping trust authority as a success criteria entirely 
though too.  The other stated error conditions are still valid though.

Jonathan G


On 11/29/18, 2:41 PM, "Rawlin Peters" <[email protected]> wrote:

    Ok, I will make sure the TP checkbox says "Validate Certificates” with
    it checked by default.
    
    However, Jonathan does bring up a good point about emitting a warning
    for authority trust validation rather than an error. My worry is that
    things in TP are typically binary: either the operation failed or it
    was a success. If we return a warning with the 200 rather than
    prohibiting it, the warning could go unnoticed leading to
    unintentional deployment of self-signed certs. With the checkbox, it
    requires that extra step of disabling validation when you are certain
    the self-signed certs are okay before adding them.
    
    Are you alright with that, Jonathan, as opposed to just allowing the
    self-signed certs to go through with a warning?
    
    - Rawlin
    On Thu, Nov 29, 2018 at 10:39 AM Gray, Jonathan
    <[email protected]> wrote:
    >
    > There are some kinds of validation that should be errors such as 
mismatched keys, broken chains, or cert ordering problems.  Authority trust 
validation (against the TO system trusts) should be a warning.
    >
    > Jonathan G
    >
    > On 11/29/18, 7:45 AM, "Jeremy Mitchell" <[email protected]> wrote:
    >
    >     +1 on not doing no double negatives...you know what i mean :)
    >
    >     On Thu, Nov 29, 2018 at 6:35 AM Eric Friedrich -X (efriedri - TRITON 
UK
    >     BIDCO LIMITED c/o Alter Domus (UK) Limited -OBO at Cisco)
    >     <[email protected]> wrote:
    >
    >     > +1
    >     > Sounds like a useful change, I know getting the right keys with the 
right
    >     > certs can be difficult.
    >     >
    >     > Is it possible to have the TP checkbox match the “polarity” of the 
API
    >     > query parameter? Rather than “Skip Validation”, can the checkbox say
    >     > “Validate Certs” and be checked by default?
    >     >
    >     > Its easier to conceptualize a positive rather than a double negative
    >     > (unchecked so dont skip validation => Do validation).
    >     >
    >     > —ERic
    >     >
    >     >
    >     > > On Nov 28, 2018, at 6:45 PM, Rawlin Peters 
<[email protected]>
    >     > wrote:
    >     > >
    >     > > Hey Traffic Controllers,
    >     > >
    >     > > If you're running a recent release of master you may find that you
    >     > > currently cannot _add_ self-signed certificates using the API 
(and by
    >     > > extension TP). However, the API still allows generating 
self-signed
    >     > > certificates. So, if your self-signed certs are generated by the 
API,
    >     > > you probably won't have any issues with those right now. However, 
if
    >     > > you're generating your self-signed certs through some other means 
than
    >     > > the API (e.g. in order to add SANs to the cert), you may find 
that you
    >     > > cannot currently _add_ those self-signed certs via the API. This 
is
    >     > > because self-signed certs do not pass the new validation in the 
_add_
    >     > > API endpoint. Since this new validation is a bit of a breaking API
    >     > > change, I'm proposing the following:
    >     > >
    >     > > 1. By default, the deliveryservices/sslkeys/add endpoint will NOT 
do
    >     > > any extra validation of the SSL cert being added. This is the old 
Perl
    >     > > behavior and has led to a lot of headaches due to it being very 
easy
    >     > > to add bad certs to a delivery service.
    >     > > 2. Add a new query parameter to this API (?validate=true) which 
when
    >     > > set to 'true' will actually perform the full validation of the
    >     > > certificate being added.
    >     > > 3. In Traffic Portal, add a checkbox next to the "Update Keys" 
button
    >     > > (which makes a request to the _add_ endpoint) that says "Skip
    >     > > certificate validation" or something. By default that checkbox 
will be
    >     > > unchecked which will add the '?validate=true' query parameter, 
meaning
    >     > > the certs will be validated. This would allow you to validate your
    >     > > certs in the API/Traffic Portal up to the point where you believe 
the
    >     > > only remaining issue is that they're self-signed. At that point 
you
    >     > > would check the box to "skip validation" to allow the addition of 
your
    >     > > self-signed certs.
    >     > >
    >     > > We really need to add validation of SSL certificates to this API
    >     > > endpoint, but at the same time I don't want this to be a breaking 
API
    >     > > change or require too much mental overhead in the UI. This would 
allow
    >     > > us to get some cert validation by default in Traffic Portal but 
still
    >     > > be able to bypass the validation for self-signed certs when 
needed. If
    >     > > using the API directly, you wouldn't need to fix anything for
    >     > > self-signed certs since the validation will not be done unless 
the new
    >     > > query param is used.
    >     > >
    >     > > Please +1 if you agree with the proposal as-is, -1 if you 
disagree or
    >     > > think the proposal needs fixing/adjusted (and please be clear on 
how I
    >     > > can change that to a +1), or just reply with a +/-0 if you don't 
care
    >     > > too strongly either way but have a different idea or some 
feedback to
    >     > > give.
    >     > >
    >     > > - Rawlin
    >     >
    >     >
    >
    >
    

  • ... Rawlin Peters
    • ... Eric Friedrich -X (efriedri - TRITON UK BIDCO LIMITED c/o Alter Domus (UK) Limited -OBO at Cisco)
      • ... Dave Neuman
      • ... Jeremy Mitchell
        • ... Gray, Jonathan
          • ... Rawlin Peters
            • ... Gray, Jonathan
              • ... Jeremy Mitchell
    • ... Hank Beatty
      • ... Rawlin Peters
        • ... Hank Beatty
          • ... Dave Neuman
          • ... Rawlin Peters
            • ... Gray, Jonathan
              • ... Dave Neuman
                • ... John Rushford

Reply via email to