Repository navigation
deSEC dynDNS: terms forbid no-op changes #8997
Replies: 1 comment 2 replies
|
Thank you for raising this! And sorry that we produced a badly behaving API client, and violated your terms of use! 😬 Since your DNS server is the authoritative source I would prefer to ask it for the stored IPs, then compare those to the results of calls to your (or some other) check-IP-services. If one of those check-IP requests fails, or equals the already stored IP, we should not overwrite the IP of that address family but This should be encapsulated in the Does that sound reasonable to you, too? Regarding generalising the no-op prevention: we didn't yet get around to implement the same support for other dyndns providers into our AIO UI, so there's no need to care for generalisation now, I would say. |
Uh oh!
There was an error while loading. Please reload this page.
Hi! deSEC here. While working on #8995 and #8996 I noticed that our dynDNS update endpoint is called every 10 minutes regardless of whether the public IP changed. That's expensive for us, which is why it conflicts with §1 of our terms of use: "only send update requests when there is something to update". As AIO has a bunch of users, this poses some risk around scaling etc.
We'd like to find a way to make the integration a well-behaved client. I can submit a PR, but I first want to figure out which way to go.
Proposal: remember the last addresses sent, and before each update ask our check-IP services (checkipv4.dedyn.io, checkipv6.dedyn.io) for the current public addresses, then only send a request when they differ; same thing on container start.
The check-IP calls are cheap (not handled by the backend but directly by nginx, no authentication database involved etc), so the 10-minute interval for that is fine. One could also compare against the live DNS, which is subject to some caching delays (but the TTL for these records in 60s, so should not be a big deal).
As I was thinking this through, I also found that your updater sends no address parameters, deSEC sets the record for the address family the container connected over and deletes the record of the other family. So AIO instances are reduced to single-stack even if a user has added an IPv6 address manually in addition to IPv4, for example. Sending explicit myipv4 and myipv6 values fixes that as part of the same change. (One should at least send the
preservevalue for the other family.)I guess before we address the no-op issue, we first have to figure out what to do about the dual-stack issue, and depending on that the no-op prevention solution need one or more state parameters.
Another question is whether the no-op prevention should be deSEC-specific, or a general behavior (which may be reasonable, but OTOH perhaps you had a reason to not do it).
What do you thinK?
All reactions