[GH-ISSUE #184] [Resolver] Allow CNAME returns on Resolver requests for Records

kerem commented

2026-03-07 22:18:40 +03:00

Owner

Originally created by @bluejekyll on GitHub (Sep 11, 2017).
Original GitHub issue: https://github.com/hickory-dns/hickory-dns/issues/184

From: github.com/bluejekyll/trust-dns@b68318e2b9 (commitcomment-24233884)

@NfNitLoop

BTW, I ran into an issue where if I do a lookup for RecordType::A, I only get back A records. It'd be super handy to get back any intermediate CNAME records too. (I want both, and would like to be able to grab them in one query instead of making two every time.) I tried reading the code to see if it was filtering those by the RecordType I passed in but I have to admit I got lost in all the async code.

Are you on the #rust IRC channel by any chance? I'm NfNitLoop there. I don't want to bloat your commit with too much conversation. :p

Originally created by @bluejekyll on GitHub (Sep 11, 2017). Original GitHub issue: https://github.com/hickory-dns/hickory-dns/issues/184 From: https://github.com/bluejekyll/trust-dns/commit/b68318e2b97701dede0702b7c9414ac41d7045f0#commitcomment-24233884 > @NfNitLoop BTW, I ran into an issue where if I do a lookup for RecordType::A, I only get back A records. It'd be super handy to get back any intermediate CNAME records too. (I want both, and would like to be able to grab them in one query instead of making two every time.) I tried reading the code to see if it was filtering those by the RecordType I passed in but I have to admit I got lost in all the async code. Are you on the #rust IRC channel by any chance? I'm NfNitLoop there. I don't want to bloat your commit with too much conversation. :p

kerem

2026-03-07 22:18:40 +03:00

closed this issue
added the
unclear

crate:resolver
labels

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@bluejekyll commented on GitHub (Sep 11, 2017):

@NfNitLoop, thanks for the feedback.

I'll start with this, the best form of feedback for me is not realtime. As I can't often spend long amounts of it in IRC or other comms channels. Though if you really want to chat directly there, I can join, we'd have to setup a shared time though.
Returning CNAMEs with the RData on A record lookups. We're getting close to the territory of crossing over from the responsibility of a resolver and the basic client. And I want to make sure the the Resolver remains easy to use, but the Client allows for anything. For reference this is the Client library: https://docs.rs/trust-dns/0.11.2/trust_dns/

That being said, maybe the Resolver should return CNAME RDatas. Can you describe your use case in a little more detail? Also, for reference, the filtering on RData type happens here in the generic lookup case:

https://github.com/bluejekyll/trust-dns/blob/master/resolver/src/lookup_state.rs#L207-L216

p.s. you might notice my comments about CNAME lookups at the top of that function.

@bluejekyll commented on GitHub (Sep 11, 2017): @NfNitLoop, thanks for the feedback. 1) I'll start with this, the best form of feedback for me is not realtime. As I can't often spend long amounts of it in IRC or other comms channels. Though if you really want to chat directly there, I can join, we'd have to setup a shared time though. 2) Returning CNAMEs with the RData on A record lookups. We're getting close to the territory of crossing over from the responsibility of a resolver and the basic client. And I want to make sure the the Resolver remains easy to use, but the Client allows for anything. For reference this is the Client library: https://docs.rs/trust-dns/0.11.2/trust_dns/ That being said, maybe the Resolver should return CNAME RDatas. Can you describe your use case in a little more detail? Also, for reference, the filtering on RData type happens here in the generic lookup case: https://github.com/bluejekyll/trust-dns/blob/master/resolver/src/lookup_state.rs#L207-L216 p.s. you might notice my comments about CNAME lookups at the top of that function.

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@NfNitLoop commented on GitHub (Sep 11, 2017):

I'm doing a (so far: proof-of-concept) rewrite of a tool that does bulk DNS queries of our customers' domains to make sure they've got CNAME/A records set up properly, so I want it to be as efficient as possible. In the C code, or using dig, making an A-record query will return both the A and CNAME records, so it's easy to see w/ one query whether a given domain was an A or CNAME.

So, when I found the Resolver::lookup(...) method I just assumed it worked like those tools. Especially since the LookupIter returns RData which can be both of those types.

I'd expect "lookup" to give me back all the data that the DNS server sent back to me as a result of that lookup.

@NfNitLoop commented on GitHub (Sep 11, 2017): I'm doing a (so far: proof-of-concept) rewrite of a tool that does bulk DNS queries of our customers' domains to make sure they've got CNAME/A records set up properly, so I want it to be as efficient as possible. In the C code, or using `dig`, making an A-record query will return both the A and CNAME records, so it's easy to see w/ one query whether a given domain was an A or CNAME. So, when I found the `Resolver::lookup(...)` method I just assumed it worked like those tools. Especially since the LookupIter returns RData which can be both of those types. I'd expect "lookup" to give me back all the data that the DNS server sent back to me as a result of that lookup.

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@NfNitLoop commented on GitHub (Sep 11, 2017):

But as you said, "as efficient as possible" may fall out of the scope of the higher-level resolver crate. If you'd decide that, it would be nice to:

document that the RData returned by lookup() will only be the type you requested.
Maybe point folks to how to get all the data in the lower-level API? (It's not clear to me. I started using the Resolver because it was easier, so I haven't learned the lower-level one sufficiently yet.)

@NfNitLoop commented on GitHub (Sep 11, 2017): But as you said, "as efficient as possible" may fall out of the scope of the higher-level resolver crate. If you'd decide that, it would be nice to: - [ ] document that the RData returned by lookup() will only be the type you requested. - [ ] Maybe point folks to how to get all the data in the lower-level API? (It's not clear to me. I started using the Resolver because it was easier, so I haven't learned the lower-level one sufficiently yet.)

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@bluejekyll commented on GitHub (Sep 11, 2017):

Yes, that makes sense. I'm just trying to limit people's exposure to the inner workings of DNS as much as possible. There are many other records generally returned as well which I wouldn't want a general purpose resolver to return, though we could debate if it should in the generic lookup function... maybe it should return all records and not limit to just the ones requested. The issue with that is that RData isn't the entire record, it doesn't return the labels (name) associated with each records RData.

For example, a DNS request will often contain in the Authority section of the Message the Nameservers to use as authorities for the domain, and then in the Additionals section it might include actual name resolutions. It varies, and it's definitely not something that I think the Resolver should return.

CNames are special of course, and I still have to workout out how to properly validate those in the Resolver. But most upstream resolvers pass back CName resolution in the Answers section, so we're getting that for free in the request. I digress a bit here though.

For what you're doing, I think you're asking for raw zone information. Have you considered doing an AXFR and just checking all the records? The TRust-DNS Client should be able to do that. You'd have to be able to connect directly to the clients NS, and they would generally need to allow you to perform AXFR requests... Then you could validate the entire zone.

p.s. thanks for the recommendation on the docs. I will definitely add that there.

@bluejekyll commented on GitHub (Sep 11, 2017): Yes, that makes sense. I'm just trying to limit people's exposure to the inner workings of DNS as much as possible. There are many other records generally returned as well which I wouldn't want a general purpose resolver to return, though we could debate if it should in the generic lookup function... maybe it should return all records and not limit to just the ones requested. The issue with that is that RData isn't the entire record, it doesn't return the labels (name) associated with each records RData. For example, a DNS request will often contain in the Authority section of the Message the Nameservers to use as authorities for the domain, and then in the Additionals section it might include actual name resolutions. It varies, and it's definitely not something that I think the Resolver should return. CNames are special of course, and I still have to workout out how to properly validate those in the Resolver. But most upstream resolvers pass back CName resolution in the Answers section, so we're getting that for free in the request. I digress a bit here though. For what you're doing, I think you're asking for raw zone information. Have you considered doing an AXFR and just checking all the records? The TRust-DNS Client should be able to do that. You'd have to be able to connect directly to the clients NS, and they would generally need to allow you to perform AXFR requests... Then you could validate the entire zone. p.s. thanks for the recommendation on the docs. I will definitely add that there.

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@NfNitLoop commented on GitHub (Sep 11, 2017):

You'd have to be able to connect directly to the clients NS, and they would generally need to allow you to perform AXFR requests... Then you could validate the entire zone.

That's a no-go. Our customers host their own DNS at a variety of providers. Even if they wanted to give us AXFR permission, managing requesting that for all of our customers would be a procedural nightmare. :p And they host many things in their DNS servers which aren't related to the sites whose DNS we care about. We're only iterating over a list of hostnames which customers have told us that they want served on our platform.

CNames are special of course, and I still have to workout out how to properly validate those in the Resolver

FWIW, because I'm only at the moment aiming for parity w/ the C tool I'm hoping to replace, I've disabled validation:

fn get_resolver() -> Result<Resolver> {
    let (config, mut opts) = read_resolv_conf("/etc/resolv.conf")?;
    opts.timeout = Duration::new(1, 0);
    opts.validate = false;
    opts.attempts = 1;
    cache_size = 100;

    return Ok(Resolver::new(config, opts)?);
}

The issue with that is that RData isn't the entire record, it doesn't return the labels (name) associated with each records RData.

Hmm, I'm not versed enough in DNS to follow here. I see the RData::CNAME(Name), though?

@NfNitLoop commented on GitHub (Sep 11, 2017): > You'd have to be able to connect directly to the clients NS, and they would generally need to allow you to perform AXFR requests... Then you could validate the entire zone. That's a no-go. Our customers host their own DNS at a variety of providers. Even if they wanted to give us AXFR permission, managing requesting that for all of our customers would be a procedural nightmare. :p And they host many things in their DNS servers which aren't related to the sites whose DNS we care about. We're only iterating over a list of hostnames which customers have told us that they want served on our platform. > CNames are special of course, and I still have to workout out how to properly validate those in the Resolver FWIW, because I'm only at the moment aiming for parity w/ the C tool I'm hoping to replace, I've disabled validation: ~~~ fn get_resolver() -> Result<Resolver> { let (config, mut opts) = read_resolv_conf("/etc/resolv.conf")?; opts.timeout = Duration::new(1, 0); opts.validate = false; opts.attempts = 1; cache_size = 100; return Ok(Resolver::new(config, opts)?); } ~~~ > The issue with that is that RData isn't the entire record, it doesn't return the labels (name) associated with each records RData. Hmm, I'm not versed enough in DNS to follow here. I see the RData::CNAME(Name), though?

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@bluejekyll commented on GitHub (Sep 11, 2017):

That's a no-go.

I figured... worth a shot :)

opts.validate = false;

That only applies to DNSSec, and you wouldn't want that, as I strip invalidly signed DNSSec response records. Though I've been considering changing that to always return, and allow the user to decide what to do with invalid records.

Hmm, I'm not versed enough in DNS to follow here. I see the RData::CNAME(Name), though?

So a response is generally broken up into three sections, Answer (records that match the query), Authority/Nameservers (SOA records and/or Nameservers for the zone) and Additional (SIG0, EDNS, etc.).

For CNAME responses you basically get this in the Answers section (though sometimes Answer + Additional), assuming this was an A record lookup:

www.example.com  CNAME host1.example.com
host1.example.com A           192.168.0.1

Taking this example, the www.example.com in the first CNAME record is the label/name of the record, CNAME the type, and host1.example.com the RData. What I was pointing out with my comment, and the reason this interface is problematic for your use case is that I don't return the entire Record, just the RData. So what this would mean is it would look like you're getting 192.168.0.1 as it's associated with a query for www.example.com and an A record type, but in fact it's associated with the name host1.example.com.

This might mean that the interface for this function should be more generalized so that this information is more obvious. But, I am still leaning toward pushing towards the client which allows for all of this data to be viewed. Does all of that make sense?

@bluejekyll commented on GitHub (Sep 11, 2017): > That's a no-go. I figured... worth a shot :) > opts.validate = false; That only applies to DNSSec, and you wouldn't want that, as I strip invalidly signed DNSSec response records. Though I've been considering changing that to always return, and allow the user to decide what to do with invalid records. > Hmm, I'm not versed enough in DNS to follow here. I see the RData::CNAME(Name), though? So a response is generally broken up into three sections, Answer (records that match the query), Authority/Nameservers (SOA records and/or Nameservers for the zone) and Additional (SIG0, EDNS, etc.). For CNAME responses you basically get this in the Answers section (though sometimes Answer + Additional), assuming this was an A record lookup: ``` www.example.com CNAME host1.example.com host1.example.com A 192.168.0.1 ``` Taking this example, the `www.example.com` in the first `CNAME` record is the label/name of the record, `CNAME` the type, and `host1.example.com` the RData. What I was pointing out with my comment, and the reason this interface is problematic for your use case is that I don't return the entire Record, just the RData. So what this would mean is it would look like you're getting `192.168.0.1` as it's associated with a query for `www.example.com` and an `A` record type, but in fact it's associated with the name `host1.example.com`. This might mean that the interface for this function should be more generalized so that this information is more obvious. But, I am still leaning toward pushing towards the client which allows for all of this data to be viewed. Does all of that make sense?

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@NfNitLoop commented on GitHub (Sep 12, 2017):

Aha, yeah, that could pose an issue for CNAME chains too. ex:

;; ANSWER SECTION:
cname-chain.nfnitloop.com. 86400 IN	CNAME	test1.nfnitloop.com.
test1.nfnitloop.com.	3600	IN	CNAME	nubo.nfnitloop.com.
nubo.nfnitloop.com.	3600	IN	A	162.243.253.171

... it would look like you're getting 192.168.0.1 as it's associated with a query for www.example.com and an A record type, but in fact it's associated with the name host1.example.com.

That's fine by me, though. When I lookup an A record for a domain, I want to know what IP address to use to connect to it. (AND, in this case, the cname(s) which got me to that IP address.) Are servers required to order things in cname1 -> cname2 -> A order when they respond? If so, the missing labels isn't that big of a deal, because I can reconstruct it by walking the RData records in order.

But, I am still leaning toward pushing towards the client which allows for all of this data to be viewed.

Yeah, it's sounding more and more like I just need to learn the client API instead. Though, one of the things that I liked about the Resolver was the ability to specify a timeout Duration. (When querying millions of hosts, I'd rather skip the slow ones than wait around for them.)

@NfNitLoop commented on GitHub (Sep 12, 2017): Aha, yeah, that could pose an issue for CNAME chains too. ex: ~~~ ;; ANSWER SECTION: cname-chain.nfnitloop.com. 86400 IN CNAME test1.nfnitloop.com. test1.nfnitloop.com. 3600 IN CNAME nubo.nfnitloop.com. nubo.nfnitloop.com. 3600 IN A 162.243.253.171 ~~~ > ... it would look like you're getting 192.168.0.1 as it's associated with a query for www.example.com and an A record type, but in fact it's associated with the name host1.example.com. That's fine by me, though. When I lookup an A record for a domain, I want to know what IP address to use to connect to it. (AND, in this case, the cname(s) which got me to that IP address.) Are servers required to order things in `cname1 -> cname2 -> A` order when they respond? If so, the missing labels isn't that big of a deal, because I can reconstruct it by walking the RData records in order. > But, I am still leaning toward pushing towards the client which allows for all of this data to be viewed. Yeah, it's sounding more and more like I just need to learn the client API instead. Though, one of the things that I liked about the Resolver was the ability to specify a timeout Duration. (When querying millions of hosts, I'd rather skip the slow ones than wait around for them.)

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@bluejekyll commented on GitHub (Sep 12, 2017):

Ah, yeah, timeouts :)

I do have timeouts on the Client, I'm just not exposing them directly in the SyncClient. If you were interested in putting a PR in for that, it should be easy to thread it through, here's the ClientFuture: https://docs.rs/trust-dns/0.11.2/trust_dns/client/struct.ClientFuture.html#method.with_timeout

The SycnClient (easier to use that the Tokio one) just creates a ClientFuture with the default timeout: https://github.com/bluejekyll/trust-dns/blob/master/client/src/client/client.rs#L371

That default is here: https://github.com/bluejekyll/trust-dns/blob/master/client/src/client/client_future.rs#L86 (5 seconds)

These docs show how to set all of this up the client: https://docs.rs/trust-dns/0.11.2/trust_dns/#setup-a-connection

There are various spots better documentation is needed, clearly.

@bluejekyll commented on GitHub (Sep 12, 2017): Ah, yeah, timeouts :) I do have timeouts on the Client, I'm just not exposing them directly in the SyncClient. If you were interested in putting a PR in for that, it should be easy to thread it through, here's the ClientFuture: https://docs.rs/trust-dns/0.11.2/trust_dns/client/struct.ClientFuture.html#method.with_timeout The SycnClient (easier to use that the Tokio one) just creates a ClientFuture with the default timeout: https://github.com/bluejekyll/trust-dns/blob/master/client/src/client/client.rs#L371 That default is here: https://github.com/bluejekyll/trust-dns/blob/master/client/src/client/client_future.rs#L86 (5 seconds) These docs show how to set all of this up the client: https://docs.rs/trust-dns/0.11.2/trust_dns/#setup-a-connection There are various spots better documentation is needed, clearly.

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@mssun commented on GitHub (May 18, 2018):

Hi, I'm facing the same issue when working on a nslookup like utility. Here is a result of www.facebook.com lookup.

$ nslookup www.facebook.com
Server:		137.189.192.3
Address:	137.189.192.3#53

Non-authoritative answer:
www.facebook.com	canonical name = star-mini.c10r.facebook.com.
Name:	star-mini.c10r.facebook.com
Address: 31.13.95.36

Also the ping tool will show the last cname (star-mini.c10r.facebook.com) when there is a cname chain and its IP address (31.13.95.36).

$ ping www.facebook.com
PING star-mini.c10r.facebook.com (31.13.95.36): 56 data bytes
64 bytes from 31.13.95.36: icmp_seq=0 ttl=58 time=83.963 ms
64 bytes from 31.13.95.36: icmp_seq=1 ttl=58 time=75.629 ms
...

Resolver cannot handle this case. Do I need the low-level client here?

PS.
I don't know why the ping util doesn't show PING www.facebook.com (31.13.95.36): 56 data bytes in the first line.

@mssun commented on GitHub (May 18, 2018): Hi, I'm facing the same issue when working on a `nslookup` like utility. Here is a result of `www.facebook.com` lookup. ``` $ nslookup www.facebook.com Server: 137.189.192.3 Address: 137.189.192.3#53 Non-authoritative answer: www.facebook.com canonical name = star-mini.c10r.facebook.com. Name: star-mini.c10r.facebook.com Address: 31.13.95.36 ``` Also the ping tool will show the last cname (star-mini.c10r.facebook.com) when there is a cname chain and its IP address (31.13.95.36). ``` $ ping www.facebook.com PING star-mini.c10r.facebook.com (31.13.95.36): 56 data bytes 64 bytes from 31.13.95.36: icmp_seq=0 ttl=58 time=83.963 ms 64 bytes from 31.13.95.36: icmp_seq=1 ttl=58 time=75.629 ms ... ``` `Resolver` cannot handle this case. Do I need the low-level `client` here? PS. I don't know why the ping util doesn't show `PING www.facebook.com (31.13.95.36): 56 data bytes` in the first line.

kerem commented

2026-03-07 22:18:42 +03:00

Author

Owner

@bluejekyll commented on GitHub (May 18, 2018):

Resolver cannot handle this case. Do I need the low-level client here?

Well, it handles CNAMEs, it just doesn't return the chain. At the moment this gets folded down into a single last item in the chain:

github.com/bluejekyll/trust-dns@a2db698afc/resolver/src/lookup_state.rs (L208-L234)

We'd need to capture all of that, without dropping it, as it does now. I'm still not sure where I stand on this ATM, but I understand what is being asked for here. An example at least for the lookup would probably be similar to what was done for SrvLookup, which has an iterator over the SRV rdatas as well as any returned IP addresses: https://docs.rs/trust-dns-resolver/0.9.0/trust_dns_resolver/lookup/struct.SrvLookup.html

If we were to support this we'd want something similar, where we'd return an iterator over the CNAME rdatas that may have led to the final IP address set, of course that may be empty...

@bluejekyll commented on GitHub (May 18, 2018): > Resolver cannot handle this case. Do I need the low-level client here? Well, it handles CNAMEs, it just doesn't return the chain. At the moment this gets folded down into a single last item in the chain: https://github.com/bluejekyll/trust-dns/blob/a2db698afc5174304f381831de3a7c7c365d8efd/resolver/src/lookup_state.rs#L208-L234 We'd need to capture all of that, without dropping it, as it does now. I'm still not sure where I stand on this ATM, but I understand what is being asked for here. An example at least for the lookup would probably be similar to what was done for SrvLookup, which has an iterator over the SRV rdatas as well as any returned IP addresses: https://docs.rs/trust-dns-resolver/0.9.0/trust_dns_resolver/lookup/struct.SrvLookup.html If we were to support this we'd want something similar, where we'd return an iterator over the CNAME rdatas that may have led to the final IP address set, of course that may be empty...

kerem commented

2026-03-07 22:18:43 +03:00

Author

Owner

@balboah commented on GitHub (Feb 18, 2020):

I'm not sure if this is related for me but when I use trust dns as a forwarder for my mac desktop, the browsers (firefox & safari) say "could not resolve hostname" for all domains which use CNAME's, for example www.amazon.com. The difference in replies from trust-dns vs original dns is that missing CNAME record

@balboah commented on GitHub (Feb 18, 2020): I'm not sure if this is related for me but when I use trust dns as a forwarder for my mac desktop, the browsers (firefox & safari) say "could not resolve hostname" for all domains which use CNAME's, for example www.amazon.com. The difference in replies from trust-dns vs original dns is that missing CNAME record

kerem commented

2026-03-07 22:18:43 +03:00

Author

Owner

@balboah commented on GitHub (Feb 18, 2020):

just realized this is a years old thread :) but this feature seems critical for operating a forwarder to support desktops

@balboah commented on GitHub (Feb 18, 2020): just realized this is a years old thread :) but this feature seems critical for operating a forwarder to support desktops

kerem commented

2026-03-07 22:18:43 +03:00

Author

Owner

@bluejekyll commented on GitHub (Feb 18, 2020):

That’s a pretty big gap you’ve identified. I had considered how this negatively impacts the forwarding resolution.

I’m not sure when I’ll have time to look at this. We’ll have to consider how to fix this. It might be that the resolver is the wrong choice in this particular case.

@bluejekyll commented on GitHub (Feb 18, 2020): That’s a pretty big gap you’ve identified. I had considered how this negatively impacts the forwarding resolution. I’m not sure when I’ll have time to look at this. We’ll have to consider how to fix this. It might be that the resolver is the wrong choice in this particular case.

kerem commented

2026-03-07 22:18:43 +03:00

Author

Owner

@balboah commented on GitHub (Feb 19, 2020):

Could you try to describe what has to be done to someone who hasn't seen .flatmap() before?
Is there a reason down the stream of calls that the CNAME records are stripped out or was it mainly because you didn't consider it to be needed by the majority of the resolver use cases?
I understand this happens somewhere around github.com/bluejekyll/trust-dns@a054d5e50a/crates/resolver/src/lookup_state.rs (L235)

My use case is to have a local forwarder which can also do the DoT/DoH lookups on behalf of the old school clients. And to have that I believe I would need to use this resolver.

@balboah commented on GitHub (Feb 19, 2020): Could you try to describe what has to be done to someone who hasn't seen `.flatmap()` before? Is there a reason down the stream of calls that the CNAME records are stripped out or was it mainly because you didn't consider it to be needed by the majority of the resolver use cases? I understand this happens somewhere around https://github.com/bluejekyll/trust-dns/blob/a054d5e50aa6a001e0966146ac8bd5f0273203fd/crates/resolver/src/lookup_state.rs#L235 My use case is to have a local forwarder which can also do the DoT/DoH lookups on behalf of the old school clients. And to have that I believe I would need to use this resolver.

kerem commented

2026-03-07 22:18:43 +03:00

Author

Owner

@bluejekyll commented on GitHub (Feb 19, 2020):

This section is fairly complex. And it has an outstanding issue with SRV target resolution. (There’s a separate issue open for that: #872).

Basically there is a first pass over the answers to see if there are an CNAMES or SRVs to resolve. If there are, it follows that chain through all the answer records. Once it hits the last see name, available, it then looks for A or AAAA records based on the request. That happens after this primary loop. If the A or AAAA isn’t available, it will issue a new request.

Right now we don’t capture the entire chain, it’s assumed only the final record is of interest. I think for the case you’ve identified, we’d need to capture all the answer and additional records into the result type. The reason I opted not to do this before is because they might be large.

Does that help answer your question? flatmap by the way is basically a combination of flatten and map. It simplifies some ownership issues by combine them. Flatten essentially exposes an inner iterator as part of the original iterator, so an iterator of iterator a is “flattened” into a single iterator.

@bluejekyll commented on GitHub (Feb 19, 2020): This section is fairly complex. And it has an outstanding issue with SRV target resolution. (There’s a separate issue open for that: #872). Basically there is a first pass over the answers to see if there are an CNAMES or SRVs to resolve. If there are, it follows that chain through all the answer records. Once it hits the last see name, available, it then looks for A or AAAA records based on the request. That happens after this primary loop. If the A or AAAA isn’t available, it will issue a new request. Right now we don’t capture the entire chain, it’s assumed only the final record is of interest. I think for the case you’ve identified, we’d need to capture all the answer and additional records into the result type. The reason I opted not to do this before is because they might be large. Does that help answer your question? flatmap by the way is basically a combination of flatten and map. It simplifies some ownership issues by combine them. Flatten essentially exposes an inner iterator as part of the original iterator, so an iterator of iterator a is “flattened” into a single iterator.

kerem referenced this issue

2026-03-16 01:55:22 +03:00

[PR #87] [MERGED] Old client to future #1235

Rows
Columns

[GH-ISSUE #184] [Resolver] Allow CNAME returns on Resolver requests for Records #87