Re: [Pki-devel] Design review request: RFC 2818 certificate compliance
On Mon, Mar 07, 2016 at 07:33:52AM +0100, Jan Cholasta wrote:
Hi,
On 29.2.2016 07:59, Fraser Tweedale wrote:
Hi all (especially those interested in certificates),
Please provide early review of my design for RFC 2818 compliance which will address the following tickets:
- #4970 Server certificate profile should always include a Subject Alternate name for the host - #5706 [RFE] Support SAN-only certificates
http://www.freeipa.org/page/V4/RFC_2818_certificate_compliance
The design is a WIP and there is no code for it yet. Looking for feedback and (hopefully) validation of the approach before committing cycles to implementing new profile components in Dogtag.
1) Do wildcard certificates need special handling? There is no mention of them in the design doc.
No special handling of wildcard certs is needed but I've added some commentary to the design page.
2) Should we accept invalid CSR where CN length is greater than 64? I wouldn't be surprised if these existed in the wild.
Good question. I agree such CSRs probably exist. There are various ways to handle them: a) Reject request (with useful message; instruction to issue SAN-only request instead) b) Issue non-compliant cert with overlong CN. It will be helpful to find out how important clients handle such certs. c) Accept the CSR but "promote" the overlong CN from CSR into a SAN dnsName, and issue a SAN-only cert. Some clients may not handle such certs very well. Personally I like (c), because the user intent is clear but we still issue a valid cert, however, I expect there are clients out there (particularly in "enterprise" environments?) that will not handle it well. I've copied pki-devel@ to solicit additional insights here :)
3) Sometimes it is not clear which parts belong to Dogtag and which to IPA itself. For example the upgrade section - I assume Dogtag should update registry.cfg and IPA caIPAserviceCert profile, but it is not clearly stated anywhere.
Thanks, I've added clarifying remarks. In brief: yes Dogtag should update registry.cfg, but FreeIPA should update the profile. Thank you for your feedback, Jan. Fraser
Hi Fraser, I think the general idea looks good. If tested to work, I actually think you should have it replace the current caServerCert.cfg and make it the default server cert profile for Dogtag. So I'd suggest you name things more generically. Just for your reference, there is an implementation that injects SAN(s) into server certs at time of Dogtag instance creation. It also allows one to put multiple SANs in one ssl server cert: https://fedorahosted.org/pki/ticket/1316#comment:14 again, it's only limited to pkispawn option so it serves a different purpose. Christina On 03/10/2016 05:06 PM, Fraser Tweedale wrote: > On Mon, Mar 07, 2016 at 07:33:52AM +0100, Jan Cholasta wrote: >> Hi, >> >> On 29.2.2016 07:59, Fraser Tweedale wrote: >>> Hi all (especially those interested in certificates), >>> >>> Please provide early review of my design for RFC 2818 compliance >>> which will address the following tickets: >>> >>> - #4970 Server certificate profile should always include a Subject Alternate name for the host >>> - #5706 [RFE] Support SAN-only certificates >>> >>> http://www.freeipa.org/page/V4/RFC_2818_certificate_compliance >>> >>> The design is a WIP and there is no code for it yet. Looking for >>> feedback and (hopefully) validation of the approach before >>> committing cycles to implementing new profile components in Dogtag. >> 1) Do wildcard certificates need special handling? There is no mention of >> them in the design doc. >> > No special handling of wildcard certs is needed but I've added some > commentary to the design page. > >> 2) Should we accept invalid CSR where CN length is greater than 64? I >> wouldn't be surprised if these existed in the wild. >> > Good question. I agree such CSRs probably exist. There are various > ways to handle them: > > a) Reject request (with useful message; instruction to issue > SAN-only request instead) > > b) Issue non-compliant cert with overlong CN. It will be helpful to > find out how important clients handle such certs. > > c) Accept the CSR but "promote" the overlong CN from CSR into a SAN > dnsName, and issue a SAN-only cert. Some clients may not handle > such certs very well. > > Personally I like (c), because the user intent is clear but we still > issue a valid cert, however, I expect there are clients out there > (particularly in "enterprise" environments?) that will not handle it > well. > > I've copied pki-devel@ to solicit additional insights here :) > >> 3) Sometimes it is not clear which parts belong to Dogtag and which to IPA >> itself. For example the upgrade section - I assume Dogtag should update >> registry.cfg and IPA caIPAserviceCert profile, but it is not clearly stated >> anywhere. >> > Thanks, I've added clarifying remarks. In brief: yes Dogtag should > update registry.cfg, but FreeIPA should update the profile. > > Thank you for your feedback, Jan. > Fraser > > _______________________________________________ > Pki-devel mailing list > Pki-devel@redhat.com > https://www.redhat.com/mailman/listinfo/pki-devel
On Fri, Mar 11, 2016 at 10:20:49AM -0800, Christina Fu wrote: > Hi Fraser, > > I think the general idea looks good. If tested to work, I actually think > you should have it replace the current caServerCert.cfg and make it the > default server cert profile for Dogtag. So I'd suggest you name things more > generically. > Thanks Christina for the feedback. W.r.t naming, can you clarify what you think should be more generic and why? > Just for your reference, there is an implementation that injects SAN(s) into > server certs at time of Dogtag instance creation. It also allows one to put > multiple SANs in one ssl server cert: > https://fedorahosted.org/pki/ticket/1316#comment:14 > again, it's only limited to pkispawn option so it serves a different > purpose. > > Christina > > On 03/10/2016 05:06 PM, Fraser Tweedale wrote: > >On Mon, Mar 07, 2016 at 07:33:52AM +0100, Jan Cholasta wrote: > >>Hi, > >> > >>On 29.2.2016 07:59, Fraser Tweedale wrote: > >>>Hi all (especially those interested in certificates), > >>> > >>>Please provide early review of my design for RFC 2818 compliance > >>>which will address the following tickets: > >>> > >>>- #4970 Server certificate profile should always include a Subject Alternate name for the host > >>>- #5706 [RFE] Support SAN-only certificates > >>> > >>>http://www.freeipa.org/page/V4/RFC_2818_certificate_compliance > >>> > >>>The design is a WIP and there is no code for it yet. Looking for > >>>feedback and (hopefully) validation of the approach before > >>>committing cycles to implementing new profile components in Dogtag. > >>1) Do wildcard certificates need special handling? There is no mention of > >>them in the design doc. > >> > >No special handling of wildcard certs is needed but I've added some > >commentary to the design page. > > > >>2) Should we accept invalid CSR where CN length is greater than 64? I > >>wouldn't be surprised if these existed in the wild. > >> > >Good question. I agree such CSRs probably exist. There are various > >ways to handle them: > > > >a) Reject request (with useful message; instruction to issue > > SAN-only request instead) > > > >b) Issue non-compliant cert with overlong CN. It will be helpful to > > find out how important clients handle such certs. > > > >c) Accept the CSR but "promote" the overlong CN from CSR into a SAN > > dnsName, and issue a SAN-only cert. Some clients may not handle > > such certs very well. > > > >Personally I like (c), because the user intent is clear but we still > >issue a valid cert, however, I expect there are clients out there > >(particularly in "enterprise" environments?) that will not handle it > >well. > > > >I've copied pki-devel@ to solicit additional insights here :) > > > >>3) Sometimes it is not clear which parts belong to Dogtag and which to IPA > >>itself. For example the upgrade section - I assume Dogtag should update > >>registry.cfg and IPA caIPAserviceCert profile, but it is not clearly stated > >>anywhere. > >> > >Thanks, I've added clarifying remarks. In brief: yes Dogtag should > >update registry.cfg, but FreeIPA should update the profile. > > > >Thank you for your feedback, Jan. > >Fraser > > > >_______________________________________________ > >Pki-devel mailing list > >Pki-devel@redhat.com > >https://www.redhat.com/mailman/listinfo/pki-devel > > -- > Manage your subscription for the Freeipa-devel mailing list: > https://www.redhat.com/mailman/listinfo/freeipa-devel > Contribute to FreeIPA: http://www.freeipa.org/page/Contribute/Code
On 03/12/2016 11:51 PM, Fraser Tweedale wrote: > On Fri, Mar 11, 2016 at 10:20:49AM -0800, Christina Fu wrote: >> Hi Fraser, >> >> I think the general idea looks good. If tested to work, I actually think >> you should have it replace the current caServerCert.cfg and make it the >> default server cert profile for Dogtag. So I'd suggest you name things more >> generically. >> > Thanks Christina for the feedback. W.r.t naming, can you clarify > what you think should be more generic and why? Actually it was more of a preemptive comment that was not specifically directed towards anything in your current design. I just took a closer look, and I think your new profile plugin name (|SubjectAltNameCopyCNDefault|) sounds good. About replacing existing caServerCert.cfg, consider keeping it, but 1. name the new profile something like caServerSANCert.cfg 2. make caServerSANCert.cfg default (enable it), and disable caServerCert.cfg by default Anyway, you get the idea. The point is that I think we should fundamentally adhere to the standard in Dogtag, so such a fix should be part of the Dogtag default. thanks, Christina > >> Just for your reference, there is an implementation that injects SAN(s) into >> server certs at time of Dogtag instance creation. It also allows one to put >> multiple SANs in one ssl server cert: >> https://fedorahosted.org/pki/ticket/1316#comment:14 >> again, it's only limited to pkispawn option so it serves a different >> purpose. >> >> Christina >> >> On 03/10/2016 05:06 PM, Fraser Tweedale wrote: >>> On Mon, Mar 07, 2016 at 07:33:52AM +0100, Jan Cholasta wrote: >>>> Hi, >>>> >>>> On 29.2.2016 07:59, Fraser Tweedale wrote: >>>>> Hi all (especially those interested in certificates), >>>>> >>>>> Please provide early review of my design for RFC 2818 compliance >>>>> which will address the following tickets: >>>>> >>>>> - #4970 Server certificate profile should always include a Subject Alternate name for the host >>>>> - #5706 [RFE] Support SAN-only certificates >>>>> >>>>> http://www.freeipa.org/page/V4/RFC_2818_certificate_compliance >>>>> >>>>> The design is a WIP and there is no code for it yet. Looking for >>>>> feedback and (hopefully) validation of the approach before >>>>> committing cycles to implementing new profile components in Dogtag. >>>> 1) Do wildcard certificates need special handling? There is no mention of >>>> them in the design doc. >>>> >>> No special handling of wildcard certs is needed but I've added some >>> commentary to the design page. >>> >>>> 2) Should we accept invalid CSR where CN length is greater than 64? I >>>> wouldn't be surprised if these existed in the wild. >>>> >>> Good question. I agree such CSRs probably exist. There are various >>> ways to handle them: >>> >>> a) Reject request (with useful message; instruction to issue >>> SAN-only request instead) >>> >>> b) Issue non-compliant cert with overlong CN. It will be helpful to >>> find out how important clients handle such certs. >>> >>> c) Accept the CSR but "promote" the overlong CN from CSR into a SAN >>> dnsName, and issue a SAN-only cert. Some clients may not handle >>> such certs very well. >>> >>> Personally I like (c), because the user intent is clear but we still >>> issue a valid cert, however, I expect there are clients out there >>> (particularly in "enterprise" environments?) that will not handle it >>> well. >>> >>> I've copied pki-devel@ to solicit additional insights here :) >>> >>>> 3) Sometimes it is not clear which parts belong to Dogtag and which to IPA >>>> itself. For example the upgrade section - I assume Dogtag should update >>>> registry.cfg and IPA caIPAserviceCert profile, but it is not clearly stated >>>> anywhere. >>>> >>> Thanks, I've added clarifying remarks. In brief: yes Dogtag should >>> update registry.cfg, but FreeIPA should update the profile. >>> >>> Thank you for your feedback, Jan. >>> Fraser >>> >>> _______________________________________________ >>> Pki-devel mailing list >>> Pki-devel@redhat.com >>> https://www.redhat.com/mailman/listinfo/pki-devel >> -- >> Manage your subscription for the Freeipa-devel mailing list: >> https://www.redhat.com/mailman/listinfo/freeipa-devel >> Contribute to FreeIPA: http://www.freeipa.org/page/Contribute/Code
On Mon, Mar 14, 2016 at 09:29:37AM -0700, Christina Fu wrote: > > > On 03/12/2016 11:51 PM, Fraser Tweedale wrote: > >On Fri, Mar 11, 2016 at 10:20:49AM -0800, Christina Fu wrote: > >>Hi Fraser, > >> > >>I think the general idea looks good. If tested to work, I actually think > >>you should have it replace the current caServerCert.cfg and make it the > >>default server cert profile for Dogtag. So I'd suggest you name things more > >>generically. > >> > >Thanks Christina for the feedback. W.r.t naming, can you clarify > >what you think should be more generic and why? > Actually it was more of a preemptive comment that was not specifically > directed towards anything in your current design. > I just took a closer look, and I think your new profile plugin name > (|SubjectAltNameCopyCNDefault|) sounds good. > > About replacing existing caServerCert.cfg, consider keeping it, but > 1. name the new profile something like caServerSANCert.cfg > 2. make caServerSANCert.cfg default (enable it), and disable > caServerCert.cfg by default > > Anyway, you get the idea. The point is that I think we should fundamentally > adhere to the standard in Dogtag, so such a fix should be part of the Dogtag > default. > > thanks, > Christina > Understood; thanks. I'll file a ticket for the Dogtag profile change. > > > >>Just for your reference, there is an implementation that injects SAN(s) into > >>server certs at time of Dogtag instance creation. It also allows one to put > >>multiple SANs in one ssl server cert: > >>https://fedorahosted.org/pki/ticket/1316#comment:14 > >>again, it's only limited to pkispawn option so it serves a different > >>purpose. > >> > >>Christina > >> > >>On 03/10/2016 05:06 PM, Fraser Tweedale wrote: > >>>On Mon, Mar 07, 2016 at 07:33:52AM +0100, Jan Cholasta wrote: > >>>>Hi, > >>>> > >>>>On 29.2.2016 07:59, Fraser Tweedale wrote: > >>>>>Hi all (especially those interested in certificates), > >>>>> > >>>>>Please provide early review of my design for RFC 2818 compliance > >>>>>which will address the following tickets: > >>>>> > >>>>>- #4970 Server certificate profile should always include a Subject Alternate name for the host > >>>>>- #5706 [RFE] Support SAN-only certificates > >>>>> > >>>>>http://www.freeipa.org/page/V4/RFC_2818_certificate_compliance > >>>>> > >>>>>The design is a WIP and there is no code for it yet. Looking for > >>>>>feedback and (hopefully) validation of the approach before > >>>>>committing cycles to implementing new profile components in Dogtag. > >>>>1) Do wildcard certificates need special handling? There is no mention of > >>>>them in the design doc. > >>>> > >>>No special handling of wildcard certs is needed but I've added some > >>>commentary to the design page. > >>> > >>>>2) Should we accept invalid CSR where CN length is greater than 64? I > >>>>wouldn't be surprised if these existed in the wild. > >>>> > >>>Good question. I agree such CSRs probably exist. There are various > >>>ways to handle them: > >>> > >>>a) Reject request (with useful message; instruction to issue > >>> SAN-only request instead) > >>> > >>>b) Issue non-compliant cert with overlong CN. It will be helpful to > >>> find out how important clients handle such certs. > >>> > >>>c) Accept the CSR but "promote" the overlong CN from CSR into a SAN > >>> dnsName, and issue a SAN-only cert. Some clients may not handle > >>> such certs very well. > >>> > >>>Personally I like (c), because the user intent is clear but we still > >>>issue a valid cert, however, I expect there are clients out there > >>>(particularly in "enterprise" environments?) that will not handle it > >>>well. > >>> > >>>I've copied pki-devel@ to solicit additional insights here :) > >>> > >>>>3) Sometimes it is not clear which parts belong to Dogtag and which to IPA > >>>>itself. For example the upgrade section - I assume Dogtag should update > >>>>registry.cfg and IPA caIPAserviceCert profile, but it is not clearly stated > >>>>anywhere. > >>>> > >>>Thanks, I've added clarifying remarks. In brief: yes Dogtag should > >>>update registry.cfg, but FreeIPA should update the profile. > >>> > >>>Thank you for your feedback, Jan. > >>>Fraser > >>> > >>>_______________________________________________ > >>>Pki-devel mailing list > >>>Pki-devel@redhat.com > >>>https://www.redhat.com/mailman/listinfo/pki-devel > >>-- > >>Manage your subscription for the Freeipa-devel mailing list: > >>https://www.redhat.com/mailman/listinfo/freeipa-devel > >>Contribute to FreeIPA: http://www.freeipa.org/page/Contribute/Code >
On Tue, Mar 15, 2016 at 07:59:57AM +1000, Fraser Tweedale wrote:
On Mon, Mar 14, 2016 at 09:29:37AM -0700, Christina Fu wrote:
On Fri, Mar 11, 2016 at 10:20:49AM -0800, Christina Fu wrote:
Hi Fraser,
I think the general idea looks good. If tested to work, I actually think you should have it replace the current caServerCert.cfg and make it the default server cert profile for Dogtag. So I'd suggest you name things more generically.
Thanks Christina for the feedback. W.r.t naming, can you clarify what you think should be more generic and why? Actually it was more of a preemptive comment that was not specifically
On 03/12/2016 11:51 PM, Fraser Tweedale wrote: directed towards anything in your current design. I just took a closer look, and I think your new profile plugin name (|SubjectAltNameCopyCNDefault|) sounds good.
About replacing existing caServerCert.cfg, consider keeping it, but 1. name the new profile something like caServerSANCert.cfg 2. make caServerSANCert.cfg default (enable it), and disable caServerCert.cfg by default
Anyway, you get the idea. The point is that I think we should fundamentally adhere to the standard in Dogtag, so such a fix should be part of the Dogtag default.
thanks, Christina
Understood; thanks. I'll file a ticket for the Dogtag profile change.
As promised: https://fedorahosted.org/pki/ticket/2233 replace caServerCert profile with one that issues RFC 2818-compliant certs Cheers, Fraser
participants (2)
-
Christina Fu -
Fraser Tweedale