Aside: Wow, we’ve hit 5-digit RFC numbers now!
I know agents are out of scope of this RFC but I love that this could easily be extended to make the JS EventSource to work on streaming AI queries.
Due to the need of bodies in requests, everyone uses POST and streaming results often use the text/event-stream protocol for responses. But this is technically a bad fit because no state is actually changing and because EventSource can only use GET for some obstinate reason. So many APIs reimplement the functionality with their own parser
> GET: Content (body) "no defined semantics"
I thought it wouldn't be a terrible idea to open up the GET method to contain a body but according to the original spec the GET body is to be ignored completely. There's also caching which would break because the important bit of the request would live in the stripped body.
Wow, it still isn’t a standard? I’ve been building with the QUERY method for years now.
I’ve enjoyed the combination with Range headers for paging, despite this tidbit:
> It is expected that these built-in features will be used instead of HTTP Range Requests
Using the QUERY request as the definition of a set, and Range to retrieve subsets seems very natural.
This makes me happy tbh, I was never a fan of creating `POST /search` endpoints when working with robust APIs
Why does this feel like GraphQL demanding everyone else solve their problems?
query strings always had size limit, seems this new type will solve it which will be really good.
QUERY has existed in spirit for nearly two decades as WebDAV's SEARCH method https://www.rfc-editor.org/rfc/rfc5323 and the thing that always killed it in practice was intermediaries. Plenty of proxies, WAFs, and load balancers either strip the body from methods they do not recognize or reject the request outright, so the guarantee that sending a body is safe evaporates the moment traffic crosses a middlebox you do not control. Until gateway and CDN support is real rather than just on paper, POST with a header marking the body as part of the cache key stays the pragmatic choice.
It's as bookmarkable as a query with its parameters in the POST data...
If this is actually going to replace GET requests w/ query strings in the wild, Im very much hoping for browser bookmarks to support keeping request parameters.
Probably won't. Probably will replace whenever POST is currently used for a query.
As "just a guy that programs" (ok, now guides agents to program) and tries to follow the rules (with a big dose of pragmatism), this totally makes sense to me. This is also the first time I've seen or heard about this coming.
I like that we now have a way to not being forced to define Resources when we want to query. It always felt like I was missing something that there could be an infinite, defined-on-the-fly number of Resources for a "part" of a given Resource. Do I really want to define "all cats that sleep more than 20 hours a day and like sunbeams and want to eat breakfast at 3 am" as a Resource? (ok, we all know that is actually the full set of cats). I'm ok that you want to define that as a Resource but in my system, it makes more sense that Cats is the Resource and I just need some accepted way to query.
I like the implementation (again, as just a guy that programs). I don't see how it could have done it better or simpler which probably hides the complexity of getting there.
I also especially appreciate how the spec is written. Opening a spec, I wonder how far I'll get before I don't know what the heck they're talking about (and, again, as just a guy that programs). I don't think it's easy to write a spec that is complete and approachable like this. Really appreciate that.
The standards have always been a bit more abstract than you use in practice. Common practice prior to this would be /catsearch?sleep=20&sunbeams=yes&breakfast=0300 where the "resource" is /catsearch and the rest are query parameters, but you could also use /catsearch/sleep/20/sunbeams/yes/breakfast/0300 which looks like a "resource", but nobody is actually enforcing that it is a "resource".
> GET request with a body was heavily considered by the IETF working group, but it was ultimately rejected in favor of creating the new QUERY method. The decision to create a distinct method came down to historical interoperability issues and strict compliance with the core architectural definitions of HTTP.
I've been sending request body along GET method for years now
> I've been sending request body along GET method for years now
Generally not a great idea. With some http implementations this is not even possible (for example, fetch)
https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API/U...
> You cannot include a body with GET requests
And transparent caching might result in weird issues.
Apparently some load balancers drop the body.
I expect all sorts of intermediaries may drop the body, since having a body is forbidden by the standard.
When it's your client talking to your server you can obviously do whatever you want - it doesn't cause problems until you want to involve third-party code, such as a reverse proxy (such as nginx) or a CDN. This includes proxies your customers may be using.
Where is it forbidden by the standard? I don't see anything in the GET definition in RFC 9110 [1] forbidding that. My understanding was that this is just undefined behavior. And not recommended due to your point about some third-party CDNs and RPs handling that UB in different ways.
Just in case anyone wants to pretend it's still that other century:
I'll forever love a long, totally plain text document like this. So many good times with video game FAQs as a kid. It really is a superior form of information in a lot of ways (not all).
I have a thesis brewing that explores how rich text WYSIWYG editors create a "what you see is all there is" cognitive bias, while plain text overcomes it.
beautiful formatting. I should crib this style template for internal work memos, it's timeless.
I don't hate it. Covers all the bases: 1.1, 2, 3/quic and solves real problems: get query limitations vs body content & post-without-mutation. Yes there are preexisting workarounds, but they're non-obvious.
Kinda pointless since traffic is encrypted. If you can terminate TLS you can apply any rule based on the content as well. Like headers which is more reliable.
It's about time.
This whole thing is non sense. It basically mixes technical constraints (body or not body) with a functional requirement that arises from people that are tied to semantics of the protocol.
HTTP is transfer protocol. It should not ever imply anything at the business level.
Yes REST made it's worst mistake out if it by giving a meaning to the verb.
Yes proxies rule how the body is re-interpreted in spite of the will of the sender (wtf).
But the original RFC states clearly that any verb can be used. This is how WebDav normalised its own.
But playing fancy by introducing a change that all HTTP implementation will have to honor is a very bad and irrational choice.
Yes this, yes that, yes the other, because proxies are in agreement with patterns are in agreement with the HTTP spec that methods exist and have semantic and functional requirements. Your 'should' seems to be discussing a hypothetical technology that is not HTTP, because HTTP has worked this way since 1997.
Just allow body for get. Problem solved.
Nah, you need the 405 Method Not Allowed to be returned by legacy systems and proxies along the way to your bleeding edge server rather than silently failing whilst dropping the params in the request body.
Use the QUERY method in your http query to query search results. Do not add query parameters.
I think the name is confusing because the term 'query' is already used to refer to http requests in general.
Just the title of the RFC confused me.
> the term 'query' is already used to refer to http requests in general
In what circles is this the case? I sometimes colloquially refer to a GET request as a query, but definitely not so on a POST, PUT or DELETE.
I'm guessing the op is referring to https://en.wikipedia.org/wiki/Query_string
Wait, it's already past 10 thousand?
Someone has an ambiguous bet predicting when RFC 10000 will be published, but the numbers went straight from 9998 to 10008. No-one wins!
https://manifold.markets/CollectedOverSpread/when-will-rfc-1...
> This question resolves to the month of publication of the lowest-numbered RFC with a number greater than or equal to 10000.
So of 10008 is the first one after 10000, that date is the one to bet on.
RFC 10000 will not be published. They're just going to skip past the number.
https://mailarchive.ietf.org/arch/msg/tools-discuss/EpoQcVt_...
RFC #s are issued sometime before publication, so they can come out out of order. I would expect 9999, 10001, etc. to show up eventually.
Everytime I think that prediction markets bets can't get worse, they do, all in weird ways. I never expected someone betting over when RFC 10,000 will be published but somehow its fits just about right for prediction markets.
just wow, people seem to be having too much money it seems for them to bet over when RFC's are gonna get released.
This isn't even one of the worst offenders on prediction market or even comparable to it but I am just amazed (in a negative manner, surprised? its just strange) by the depth on what people actually bet on these markets.
People aren't betting real money on this. Manifold uses "mana" points similar to HN karma, which is why you get more for-fun silly bets. I don't see anything inherently wrong with it. Disclosure: my mana net worth is 75k; I haven't been active on Manifold.
Ah okay. I didn't know that.
Interesting thing actually. Seems similar to the trend in South Korea recently where you can online shop to get the thrill of shopping but you aren't actually paying with money.
But I am unsure of the overlap between manifold and polymarket/kalshi. I imagine that some might win in manifold and try to bet on polymarket to win "real" money which ends up being a bit gambling-esque.
But good for manifold for atleast not playing with real money but rather points like this. I would argue that Manifold might be better than polygon/kalshi in terms of net positive outcome of its existence for the world perhaps.
Interesting trend. I have found a short article about it: https://www.odditycentral.com/news/south-koreas-fake-online-.... It seems nothing long-form has been written about it in English yet.
There is an overlap between Manifold and Polymarket/Kalshi. At the very least, Polymarket is more liquid, which creates opportunities for arbitrage and incentives for Manifold users to follow Polymarket. There is something at stake on Manifold itself if you choose to pursue it. There have been ways to convert mana to charitable donations (to your preferred charity), tickets to Manifest (the Manifold conference), and also merch and now prize drawings. Mana is like HN karma in that being at the top gives you status and bragging rights and suggests technical competence.
Why not standardize a body in the GET request (which isn't forbidden per spec and works in many places already but isn't supported everywhere because it's not mandated to support it)?
Too many servers ignore/drop/reject body in GET requests. RFC 9110 does allow it, but is only recommended if server documentation states that it is supported.
These servers will likely also reject the QUERY request though…
But at least in that case you may get a much more meaningful 405 Method Not Allowed response rather than the server just silently dropping the GET body content.
I wonder if HTML forms will add support for QUERY:
<form action="..." method="query">
Supporting more than GET/POST in HTML forms has been my dream for decades. There's a WHATWG proposal to do just that if you want to add your voice: https://github.com/whatwg/html/pull/11347
Depends whether your form submission should expect side effects or not. Most forms submissions have side effects. If the effect is truly idempotent, wouldn't PUT be a better verb? That is also supposed to be idempotent.
You can’t use PUT as a form action though.
GET and QUERY are both idempotent.
Now HN’s UX can finally be decent.
The team will have to wait for a new header and textarea specs to fix the rest of the jank.
This site is so awful lol. Why don’t they update it?
Where does HN use POST for safe operations? I can't think of any.
Comment submission isn't safe, so QUERY can't be used there. And it doesn't suffer from the problem anyways, since HN returns a 3XX on successful submission, so refreshing doesn't show a warning.
This is better solved with the post redirect get pattern.
That is the good old fashion workaround. But why is it better than a form causing an HTTP QUERY.
If we can do QUERY forms, it would be an ideal time to add JSON encoding for forms.
One oddity of forms: the result of a form POST is a page that has a location (the URL) but that cannot loaded via that location. As far as I know, the fact that the page is a POST and not a GET is not stored anywhere visible to the user or to JS. And refresh works oddly.
If method=QUERY were added, there would be a new variety of this weirdness.
At least browsers wouldn't have to warn users that they'd be resubmitting data if they reload the page after submitting a query form, since query requests are intended to be idempotent
You still get the nastiness that the Sec-Fetch-* state gets mostly trashed when you hit refresh. And someone would need to figure out how CORS preflight interacts with refresh, which is not currently an issue with POST. (The current "simple request" behavior or whatever it's called is a real mess and is the cause of a lot of CSRF vulnerabilities.)
Forms, HTTP implementations, public API surfaces, and all for what exactly. Introducing a new verb for this feels profoundly misplaced
At least support - or lack thereof - for a new verb is unambiguous (compared to changing the semantics of GET)
Idempotency is an important attribute for correctness. Yep, you can document that POSTing to $ENDPOINT is idempotent, but you can't communicate that to caching layers throughout the network. QUERY, by definition, is idempotent and cacheable.
Great point. I wish more people realized that intuitively.
Does anyone know if this approach works at significantly larger scales? Curious about where it breaks down.
Larger scales like what? I expect that everywhere you currently cache GETs you can cache QUERYs. But does caching GETs work at scale?
Including a strong motivating example might have helped sell this, using an example that could trivially be expressed as a GET is extremely distracting.
Even imagining a QUERY with a large JSON filtering structure, or say an image input as request body, it feels extremely odd to include the request body as part of the cache key. It also implies an unbounded and user-controlled cache key, with the only really meaningful general caching strategy being bitwise compare of the request body (or a hash), which in a hostile scenario implies cache busting would be trivial.
This invokes multiple semantic oddities in one go with obvious difficulties for a very niche use case. If I'm writing a service that needs complex filtering or complex input like an image, any form of caching (e.g. individual data columns of a join, or embeddings keyed by perceptual hashes of a decoded image input) is going to be far away from the HTTP layer and certainly unrelated to the exact bit representation of the request on the wire.
Why even bother trying to capture this in a generic way?
I would be far more inclined to try and capture this caching semantic as a new header for POST. Something like "Vary: request-body" or similar. Perfectly backwards compatible and perfectly ignorable for all but the 0.1% of CDN use cases where the behaviour might turn out useful
> It also implies an unbounded and user-controlled cache key.
While the concern is valid, caching is entirely optional at query level, therefore it is totally valid to cache only certain "filters".
One example - I'm building an MCP server at the moment for a database I'm working on. In ChatGPT I want to do dry-run posts first that roll back before committing - both are POST requests with a property - and it loves to trigger the safety layer in the tools (for various reasons, it's hard to debug exact causes)
But I think this would make it better - QUERY before POST means different request types, not just the same with a safety flag.
> Why even bother trying to capture this in a generic way?
I guess it's about resolving the odd semantics of using POST which is not idempotent and thus allowing easier control flow of caches and retrys.
Your perspective is 100% correct if you think at the application-layer, but with a dedicated method, you can have that behaviour out-of-the-box out of your HTTP infrastructure (whether it's at your hyperscaler's router or your apache/nginx/browser whatever) and stop implementing yourself the post-as-a-query edge case.
If you control the full stack then the functionality described here can be implemented with POST. The only way this comes into play is if some second party client of your service is trying to impose rules on how your backend works. My answer to that is no. I will be defining the contract by which my services operate.
The browser can simply store a collision resistant hash (e.g. SHA-256) of the body, if it wants a smaller cache key. I can't really think of any caching related attacks that don't equally apply to a query parameter. Generating a unique 30 character query parameter is just as easy as generating a 30 MB request body, if you want to flood the cache.
Not necessarily that simple, as you'd have sort all the input parameters to maintain a useable cache key. Not especially difficult, but if the data is large and so re-allocation and sorting is required, then you're starting to open up the attack surface where bugs might have been introduced.
> It also implies an unbounded and user-controlled cache key,
The query part of GET's URI is also barely bounded in practice and user-controlled, and is indeed used as part of the cache key (because it's a part of URI), so I am not sure why you raise this objection at all.
> and user-controlled
I've found some sites that tack on a session ID and if you try to tamper with the URL in any way, it sends you back to "Page 1" really annoys me lol at that point let me skip to any page with your web UI.
Well, because it is more code. Current caching software caches by headers + query string. It now needs to be expaned to cache by body too.
It feels very pointless and there is no drawback of just using POST
Is caching not the primary reason to use this over POST? You should never want to cache POST requests.
There is: your browser or other type of client does not know it can repeat a POST request if it fails, whereas a QUERY request can be freely repeated in case of errors.
Not all usage scenarios are the public internet, and something doesn't have to be useful on the public internet to be standardized.
Realistically, systems for the public internet will use a secure hash as the cache key so it'll always be the same size. The cache key already includes a URL that can be very long, and an arbitrary set of header values.
Except that by definition, in a URL the data has no implicit meaning so for a cache hit you need an exact match, including order and case, but for a list of POST parameters, they could legitimately be in any order and so you can't just hash it all as a blob, you need to sort the keys, possibly copy data around (unless using keys plus hash), probably allocating more memory, etc. I'm pretty certain we'll see at least one CVE out of the first few implementations of this!
POST/QUERY data can be in any format. Who are you to say order doesn't matter? Are you sure you can even parse it? Mine is in DES-encrypted (with key "password") base85 DER, you really gonna implement that in your proxy?
I would use a hash of the body content (the query) as a URL parameter
/?hash=123456789
Why? That's pushing more work to do both on yourself and the cache.
Sure you can provide an image as request body, but you could already do it with b64 query parameter. If you try hard enough, you can poorly use any proposed standard. GET with query parameters already is opaque and makes cache busting trivial.
Query parameters are length-limited, because HTTP URIs are: https://www.rfc-editor.org/info/rfc9110/#section-4.1-5. There is no expectation for arbitrarily long HTTP URLs to be functioning.
Your link doesn't say URIs are length-limited
I'm guessing you never hit this issue then, but it's a real issue. Whether or not it's in the RFC as a hard limit it doesn't matter, no HTTP server will allow unlimited sized URIs.
You simply can't base64 large payloads and you're stuck with workarounds.
You are guessing wrong. Thanks, I know specific implementation will come with their limits. This will equally apply to QUERY body size and caching strategy.
Are we seriously ok with linking the RFC as source while providing a statement that doesn't match? RFC does matter.
The RFC does say "It is RECOMMENDED that all senders and recipients support, at a minimum, URIs with lengths of 8000 octets in protocol elements."
One can infer from the RFC that you can reasonably expect many implementations to fail beyond 8000 characters, and that there are no guarantees up to that either.
True, the RFC doesn't specify a limit, but it does clearly indicate that it's not unbounded, nor should you expect it to be.
Why not just define the semantics of a GET request body?
There's countless proxies in the wild that would not behave correctly with an RFC-defined GET-with-body, and there's no way for a client to know if that's the case.
QUERY has the advantage of getting default behaviour from most proxies (which at least is well behaved even if inefficient). If there are any proxies that just drop QUERY requests, at least they won't silently mangle the request.
This is the same way that instead of improving how HTTP 301 was specified, HTTP 308 was created. It's a pragmatic move.
Proxies often delete it
They could be updated to not delete it, like they would require for this new method anyway.
Agree.
They should not delete the body in the first place.
The HTTP standard does not define a body for GET requests. Therefore, proxy implementations will typically only copy the data from the request header when sending it off to its next destination. They don’t technically “delete” anything. This saves compute (and possibly bandwidth, if the request happened to have a body, although the whole point is that one can safely assume a GET request does not have a body per the standard).
If we're going to play the "should" game, whatever originated the body "shouldn't" have because the spec says that's illegal.
Although we could also go with, a proxy shouldn't delete the body, it "should" reject the request outright as ill-formed.
The meta-point I'm making here, which I'm sure will be missed if I don't spell it out, is that if we're going to talk about what "should" be done when it is explicily out of the scope of a standard, there's no way around the fact that there are multiple completely sensible ways to extend the standard and there's every reason to expect that in the real world people aren't going to agree. Sometimes they manage to, but even then often quite imperfectly. Our human intuitions that standards are something that are "built" is perhaps not wrong, but you can also productively look at standards as taking the raw material of all possible things two systems could send to each other and removing possibilities. If you reach into a space that has been explicitly removed, you can't expect everyone else to do so in exactly the same way you will.
Wouldn’t just putting an etag on POST requests accomplish the same thing? If I’m understanding it the server has to maintain state to ensure idempotency.
Unlike POST, however, the method is explicitly safe and idempotent, allowing
functions like caching and automatic retries to operate.
QUERY is GET with a request body. So it must be safe, not just idempotent. Where safe means it has no significant side-effects. Typically servers will not keep any state for QUERY requests.
There is one interesting variant though, which uses state: The client sends a QUERY containing the full query, and the server returns a url usable with GET with which this query can be triggered in the future. Similar to prepared statements in SQL databases.
Using QUERY for GraphQL queries (not mutations) would be a good match. These only read data, but are sometimes bigger than the url length limit.
Ideally, libraries like FastAPI, etc. could be configured to translate QUERYs to GETs, until you can rewrite your code to automatically support both.
Interestingly, despite the QUERY request being safe, the RFC says it's subject to preflight requests:
> A QUERY request from user agents implementing Cross-Origin Resource Sharing (CORS) will require a "preflight" request, as QUERY does not belong to the set of CORS-safelisted methods (see [FETCH]).
That paragraph merely describes how existing browsers behave, it doesn't specify how future browsers must behave. After all, a HTTP RFC isn't really the right place to specify browser specific behavior like CORS, that belongs in a W3C/WHATWG specification.
Thanks for the explanation!
I still don’t get how idempotency can typically be ensured without state. It very much depends on data model and application design. Even side effects like using a user’s lookup quota need to be handled at a higher layer than HTTP (I think?).
> I still don’t get how idempotency can typically be ensured without state.
Well, how is "GET /index.html HTTP/1.1" made idempotent in practice without (additional) state?
Imagine a forum where comment ids are client-generated UUIDs, and comments are inserted with "ON CONFLICT IGNORE". Submitting the same comment twice would simply be a noop
But what the Query method really targets are things like a graphql query that can be multiple kb for a single query, but only reads data. Sure, it might count against rate limits, trigger logs, etc. But at a conceptual level resubmitting the same query should give the same result (if the data didn't change). And since you are only reading data, resubmitting is safe
Minor side-effects like quotas or request logging are generally ignored when considering the semantics of http methods. I don't see any complications for QUERY that don't already apply to GET. It just allows you to bypass the url length limit by putting the data in the body instead of the url itself.
Exactly. "GET / is idempotent" ... "But I just launched a DDOS against / and now it returns 502...
Yes it varies. Using the QUERY method doesn't automatically mean your app is idempotent - it means the browser, and any intermediaries, can assume it's idempotent. So when you go forward and then back they're free to reissue the request and you won't get the "this may repeat whatever you just did" popup.
If it's not actually idempotent but you're telling the browser it is, of course you may cause bugs. Same as GET.
I think PUT, which is idempotent, would still require the "repeat action" warning, since it will overwrite any changes that happened to the same resource in the mean time (unless used with `if-match` or similar). QUERY won't require such warnings because it's safe, not just idempotent.