search

getdnsapi / stubby

Stubby is the name given to a mode of using getdns which enables it to act as a local DNS Privacy stub resolver (using DNS-over-TLS).
https://dnsprivacy.org/dns_privacy_daemon_-_stubby/
BSD 3-Clause "New" or "Revised" License
1.2k stars 100 forks source link

Fill the gap with Unbound functionality #340

Open wellloaded opened 1 year ago

wellloaded commented 1 year ago

Hello Sara, I came across this comparison table between Stubby and Unbound. I wasn't aware of this but it seems like there are indeed some gaps in terms of functionality that could (should?) be filled by Stubby:

Criteria | Stubby | Unbound -- | -- | -- License | BSD 3-Clause | BSD 3-Clause Implementation language | C | C DNS-over-TLS (DoT) | Yes | Yes DNS-over-HTTPS (DoH) | Yes | Yes DNS-over-QUIC | Yes | Yes DNSSEC Validation | Yes | Yes Cache | **No** | Yes Recursive queries | **No** | Yes Forwarding | Yes | Yes Load Balancing | **No** | Yes Multiple Root Support | **No** | Yes Round-robin DNS | **No** | Yes Response Rate Limiting | **No** | Yes API | Yes | Yes Operating Systems supported | Linux, macOS, Windows | Linux, macOS, Windows

HTH Thanks

ArchangeGabriel commented 1 year ago

The consensus here is that you should run both together: Unbound as your local DNS resolver, and Stubby as its upstream to relay requests with Do to selected upstreams (because it at least used to be better in this field). Though that looking at your table, I might be in need to update my Unbound knowledge if that one supports all DNS-over- as well as Round-robin (something that was requested for Stubby).

saradickinson commented 1 year ago

Hi there,

Can you point to where you found this as I think it needs some correction/clarification. Some of the criteria only make sense for Unbound when acting as a recursive resolver (e.g Load balancing, Multiple root support, Response rate limiting.....)

Firstly, Unbound is a validating, recursive, caching DNS resolver that can also be configured as a forwarding proxy.

In both scenarios it can do DNSSEC validation and caching.

Stubby (as packaged) is just a validating, forwarding proxy (hence no cache, no recursion).

Historically Stubby had support for forwarding over DoT quite a while before Unbound and also had many more configuration options to better manage the DoT connections to upstream resolvers. Unbound now supports basic DoT forwarding, but still doesn't have all the configuration that Stubby does.

So (as @ArchangeGabriel said) what many folks do is run Unbound locally as a forwarder to provide local DNSSEC validation with a cache, and then queries from Unbound are sent via Stubby which forwards them to a recursive over DoT.

Example config for this is here Implementation status of privacy related features is here

Some further points:

ArchangeGabriel commented 10 months ago

@saradickinson Doing some triaging in my GitHub notifications emails I reviewed https://github.com/getdnsapi/stubby/issues/330 and with this too, I think that Stubby binding to 53 by default is not ideal. It makes people think that Stubby should be used on its own while it does not do caching.

Maybe having a different default port and insisting in comments/documentation that a caching recursive resolver should be used in front would be better. And would allow for better security with PrivateUsers by default.

But anyway, Stubby and getdns do not seem to get much traction these days, so…

wellloaded commented 9 months ago

Hi there,

Can you point to where you found this as I think it needs some correction/clarification. Some of the criteria only make sense for Unbound when acting as a recursive resolver (e.g Load balancing, Multiple root support, Response rate limiting.....)

Apologies for the long delay in the answer. I'm uncertain on the actual source/investigation but this was picked up from the tomato forum: https://www.linksysinfo.org/index.php?threads/tomato-unbound.77832/post-341744