 |
 |
PS: If one single server (or group of identical servers, horizontally scaled) needs to respond to multiple hostnames, I would use a single SAN cert with multiple hostnames. If multiple entirely different servers just happen to be different *. university.edu -- I would not use a SAN cert or a wildcard cert, I would give them each their own separate cert. This is, IMO, more contemporary/"correct" operations -- one cert per server build, not one cert per organization. The reasons people used one cert per organization were mainly about cost (of providers that charge you per cert), or with management difficulties, both things letsencrypt is specifically meant to deal with to make one cert per build possible. The advantages of one cert per build are security partitioning, one machine being compromised has not compromised the SSL certs for your entire organization. On Mon, Jun 19, 2017 at 3:00 PM, Jonathan Rochkind <[log in to unmask]> wrote: > There's no reason you _need_ to use a wildcard cert for many hosts. You > can use a separate cert for each. The reason people prefer a wildcard cert > is because it was a pain to _get_ and keep track of all those certs. > > letsencrypt archicture encourages you to just do that. The certs are > automatically obtained and automatically renewed, there's no reason you > need the same cert accross multiple hosts, each host gets it's own cert. > (Which also means if one of them gets compromised, and you need to revoke a > cert, you just need to revoke one host's cert, not a wildcard cert applying > to all of them). (And yes, automatic renewals are not hard with > letsencrypt, it's specifically intended you do automated renewals, there > are a variety of software and scripts for different environments to do it). > > I don't see anything wrong with that, really. > > There are cases where you really do need a wildcard cert -- an app which > has _dynamic_ hostnames, like a hostname for each user account (eg > jrochkind.github.io). letsencrypt isn't going to work there, you really > do need a wildcard cert. > > But just for a lot of hosts on the same TLD? They don't need a wildcard > cert, and there are reasons to prefer them _not_ having a single wildcard > cert (the revocation case, especially if they are all administered by > different units), they can each just have their own cert. > > I am not sure what Kyle means by "encryption hides attacks". Personally, > I think SSL encryption is a requirement for contemporary professionally > managed websites. But the question of whether to use https or not at all > -- is really a separate issue than whether to use letsencrypt/acme for > your SSL certs. If you decide you don't want/need https/SSL encryption at > all, then you don't need to consider letsencrypt as opposed to a more > manual cert provider at all. :) > > I don't see any real reason letsencrypt would not be viable for a library. > You do need to have enough sysadmin ability to set up the automatic > renewals, and understand what's going on, yes, that could be a barrier I > suppose. The main potential barrier I see is letsencrypt rate limits on > hosts-per-tld, for an academic institution that is going to have > hundreds/thousands of hosts within the TLD (*.university.edu). It seems > they will exempt a university from these rate limits with an email request. > > > I feel like there's a lot of FUD about letsencypt going around for some > reason. > > I believe there are lots of all letsencrypt certs granted somewhere, it > might be possible to find those and look for *.edu's to find peer > institutions. > > Jonathan > > > > On Mon, Jun 19, 2017 at 2:00 PM, Kyle Banerjee <[log in to unmask]> > wrote: > >> I almost wrote it wouldn't work, but what works always depends on the >> particulars of your situation. For example, depending on how many domains >> you need and what mechanisms you're using, you might be able to use >> Subject >> Alternative Name (SAN) certificates to mitigate the lack of a wildcard >> certificate. Another thing I was thinking about as I wrote that is that a >> growing number of libraries provision resources with vendors such as >> Amazon >> -- for that, you'll need the cooperation of your institution. >> >> Automating renewal is a good practice. Remember when when the doi.org >> cert >> expired a few years back? Wasn't pretty and could have worked out much >> worse had the domain squatters been on the ball. It's not hard to >> automate, >> and instructions are easy enough to find. Even when squatters aren't an >> issue, expired certs cause all kinds of scary warnings. >> >> One of the big problems libraries face is that a lot of free stuff is not >> viable for many libraries that need help the most. The whole problem is >> these institutions often lack both staff and technical resources. And even >> if they do have someone with the requisite skills to build great stuff out >> of virtually nothing, they risk serious problems when that person leaves >> and they can't replace them with someone with similar abilities. >> >> It is taken as gospel here that encryption is always good, but it's always >> important to be aware of tradeoffs. For example, encryption hides >> attacks. It can instill a false sense of security -- there are lots of >> ways to track activity that aren't affected by encryption. It prevents >> caching and complicates complying with CIPA as well as state law filtering >> requirements, and it could create issues if your services must communicate >> with legacy apps. >> >> kyle >> >> >> >> On Mon, Jun 19, 2017 at 7:54 AM, Kyle Breneman <[log in to unmask]> >> wrote: >> >> > Thanks for chiming in, Kyle. I think, in your second-to-last sentence, >> you >> > were about to say "impossible." Is that right? Also is it difficult to >> > setup automatic certificate renewal? For the record, I'm not trying to >> > bypass any organizational processes here, just doing some legwork in >> hopes >> > of handing campus IT a suggestion that will save them money. >> > >> > Kyle >> > >> > On Mon, Jun 19, 2017 at 9:51 AM, Kyle Banerjee <[log in to unmask] >> > >> > wrote: >> > >> > > There are a few other catches. For example, you need to be able to >> run an >> > > appropriate ACME client and set up automatic certificate renewal since >> > the >> > > maximum length you can get is 90 days. You also can't get wildcard >> > > certificates which makes doing things like proxying by host name (e.g. >> > > ezproxy). Your organization might also care if you bypass their >> process >> > for >> > > getting domain names. >> > > >> > > kyle >> > > >> > > On Mon, Jun 19, 2017 at 5:41 AM, Jonathan Rochkind <[log in to unmask] >> > >> > > wrote: >> > > >> > > > Here's a thread about per-TLD rate limits being a problem for >> > > universities; >> > > > it seems per a post at the end of that thread that letsencrypt might >> > > exempt >> > > > your institution from ratelimits, but an official agent of the >> > university >> > > > needs to submit the request: >> > > > >> > > > https://community.letsencrypt.org/t/rate-limiting-at-an- >> > > > educational-institution/5910/24 >> > > > >> > > > >> > > > >> > > > On Mon, Jun 19, 2017 at 8:27 AM, Kyle Breneman < >> > [log in to unmask]> >> > > > wrote: >> > > > >> > > > > Thanks for that detailed and interesting reply, Jonathan. >> > > > > >> > > > > On Sun, Jun 18, 2017 at 12:35 PM, Jonathan Rochkind < >> > [log in to unmask] >> > > > >> > > > > wrote: >> > > > > >> > > > > > Just to clarify, by "Commercial certificates offer stronger >> proof >> > of >> > > > > > identity", you mean an "Extended Validation" (EV) certificate. >> > > > > > https://en.wikipedia.org/wiki/Extended_Validation_Certificate >> > > > > > >> > > > > > If you are getting a 'commercial certificate' that is a standard >> > > > 'domain >> > > > > > validated' cert instead of an EV cert, you are not getting any >> > > stronger >> > > > > > proof of identity than you would from letsencrypt. >> > > > > > >> > > > > > The cert used at https://www.ubalt.edu does NOT appear to be >> an EV >> > > > cert, >> > > > > > but an ordinary domain validated one. (Additionally, that >> > particular >> > > > web >> > > > > > page serves http: images , triggering browser mixed content >> > > warnings!). >> > > > > > >> > > > > > Same thing for the cert at https://langsdale.ubalt.edu/. >> > > > > > >> > > > > > Looking at another Maryland public university: >> https://umd.edu/ >> > > > appears >> > > > > > similar. NOT an EV cert, and additionally serving http assets >> > > > triggering >> > > > > a >> > > > > > mixed content warning. >> > > > > > >> > > > > > I'm actually having trouble finding an academic institution, or >> > even >> > > a >> > > > > > standard ecommerce site, that DOES use an EV cert. >> > > > > > >> > > > > > You can tell it's an EV cert when chrome or Firefox put the >> name of >> > > the >> > > > > > organization in the location bar to the left of URL. >> Additionally, >> > > in >> > > > > > Firefox, if you click that name, then click the right-chevron >> 'more >> > > > info' >> > > > > > icon, then click "More information", under "Website Identity" it >> > will >> > > > > list >> > > > > > an "Owner:" for an EV cert. For an ordinary domain-validated >> cert, >> > it >> > > > > will >> > > > > > list "This website does not supply ownership information" >> instead. >> > > > > > >> > > > > > Here's an example of an EV cert, the cert on digicert.com, a >> > seller >> > > of >> > > > > > certs: >> > > > > > >> > > > > > https://www.digicert.com/ >> > > > > > >> > > > > > If your cert is not EV but is just "domain validated", then >> despite >> > > it >> > > > > > being "commercial" it supplies the same level of proof of >> identity >> > > as a >> > > > > > letsencrypt cert -- proof of control of the domain at the time >> the >> > > cert >> > > > > was >> > > > > > issued, either way. >> > > > > > >> > > > > > >> > > > > > >> > > > > > On Sat, Jun 17, 2017 at 1:53 PM, Cary Gordon < >> [log in to unmask] >> > > >> > > > > wrote: >> > > > > > >> > > > > > > We are starting to roll out LetsEncrypt for all of our >> services >> > and >> > > > > > > clients who do not use or want commercial certificates. >> > > > > > > >> > > > > > > Note that LetsEncrypt offers only domain authentication, in >> most >> > > > cases >> > > > > > > specifically validated by your control of the server. >> Commercial >> > > > > > > certificates offer stronger proof of identity. >> > > > > > > >> > > > > > > We recommend commercial certificates for any sites that >> conduct >> > > > > financial >> > > > > > > transactions or require HIPPA compliance. >> > > > > > > >> > > > > > > Thanks, >> > > > > > > >> > > > > > > Cary >> > > > > > > >> > > > > > > Cary Gordon >> > > > > > > The Cherry Hill Company >> > > > > > > http://chillco.com >> > > > > > > >> > > > > > > >> > > > > > > > On Jun 16, 2017, at 12:34 PM, Kyle Breneman (via lita-l >> Mailing >> > > > > List) < >> > > > > > > [log in to unmask]> wrote: >> > > > > > > > >> > > > > > > > Apologies for cross-posting... >> > > > > > > > >> > > > > > > > Anyone out there working at a public institution that's >> using >> > > Let's >> > > > > > > Encrypt for security certificates? I just suggested to our >> > campus >> > > IT >> > > > > > that >> > > > > > > we switch to using Let's Encrypt. They told me it would need >> to >> > > > clear >> > > > > > > State of Maryland approval process first, and suggested that >> it >> > > would >> > > > > be >> > > > > > > helpful to be able to point to other public institutions that >> are >> > > > using >> > > > > > it. >> > > > > > > > >> > > > > > > > Regards, >> > > > > > > > Kyle Breneman >> > > > > > > > Integrated Digital Services Librarian >> > > > > > > > University of Baltimore >> > > > > > > > >> > > > > > > > To maximize your use of LITA-L or to unsubscribe, see >> > > > > > > http://www.ala.org/lita/involve/email >> > > > > > > >> > > > > > >> > > > > >> > > > >> > > >> > >> > >