+1 to Option 2 for me. >From what I remember of the meeting (it will be nice to review once the recording is posted to make sure we didn't forget any important viewpoints), the major concern wasn't just difficulty of client-side implementations of UUIDv7, but two aspects of being risky to allow underspecification:
1. In the same vein as what Yun said here, some people expressed a preference for an explicit separate field to be more explicit about intent and semantics instead of relying on the embedded portion of the UUIDv7 2. There was concern about letting servers have divergent *behaviors* that depend on interpreting the UUIDv7 in ways not prescribed in the spec For the second concern, my take on it is that if we're really going to intend for servers to use it in conjunction with their idempotency-key-expiration window to reject old requests, we should go all the way and really specify the time-range rejection behaviors very explicitly. Otherwise, overspecifying half of the protocol (UUIDv7) while underspecifying the server-side behaviors could lead to problems down the road. There was a comment in the PR discussion ( https://github.com/apache/iceberg/pull/14196#discussion_r2453115088) which I think is worth discussing further here: if a client is misbehaving but assume things are fine because one server is > not complaining, but things start to go wrong when the client is addressing > a different server (like rejecting a non v7 uuid or an old one), then it's > a major client issue. IMHO server is not required to do anything (it's not even required to > support idempotency key) but if it chooses to, it should be able to > leverage as much information the key can give. The problem here is that if the spec is underspecified, then it's not clearly defined whether the client is "misbehaving" in those underspecified scenarios - for example, is the client misbehaving if their clock skews from the server by 1 second? 10 seconds? 60 seconds? 300 seconds? Is it misbehaving if its time appears to be "in the future" even after accounting for an allowed clock skew? If the client doesn't know whether intermediate proxies are retrying, can it *rely* on the server to reject requests older than the idempotency-key-lifetime? Or does it still need to proactively guarantee no retries older than the idempotency key lifetime? The goal of the spec and consistent server implementations is to help ensure a good experience for the client-side, so it immediately knows whether it's misbehaving. I disagree with allowing the server to leverage as much information as the key can give when it comes to leveraging information for actual behavioral decisions. It's one thing to choose not to implement an optional feature, and advertise the server capabilities as such, but another thing for idiosyncratic behaviors to start arising when a feature *is* implemented. It's certainly true that it's *possible* to leave it unspecified and let various de-facto standards emerge organically, but that seems to be the kind of divergence we're precisely trying to prevent by having this shared REST spec, just like the stricter web standards enforcement in the early 2000s/2010s to make cross-browser development manageable. A thought experiment - if we were to specify UUIDv1 as required and leave the usage of the internals of UUIDv1 underspecified, would it be appropriate for a server to choose to use the MAC-address portion (bits 80 through 127) for special behavioral handling that draws conclusion about multiple requests coming from the (allegedly) same client machine? For example, one could imagine fancy heuristics where one decides to reduce an idempotency-key retention window after consecutive requests from the same client machine indicate that the network connection from that machine is "healthy". Overall, it seems like getting the lifecycle-window handling fully thought through will require some work, and for such a complex piece it seems cleaner to have separation of concerns here, with the UUID in itself just serving the purpose of being a UUID and a separate explicit timestamp field allowing better reasoning about protocol expectations, or lack thereof (e.g. a client that doesn't have a reliable clock, but is only worried about intermediate proxies performing retries can still be successful in using UUIDv4 without giving the explicit idempotency-key-timestamp, if intermediate proxies can only cause retries for a duration much shorter than the key lifetime). On Wed, Oct 29, 2025 at 8:05 PM Steven Wu <[email protected]> wrote: > > When the server decides to expire a key, will it rely on the > server-receive timestamp, the client request timestamp, or the key creation > timestamp? > > Yun has a good point here. Current spec PR for "idempotency-key-lifetime" > should be "server-receive timestamp". > > > Now, let's say a client is non-conformant and is reusing the same key or > is not applying correctly the lifetime directive, but send the same request > to the server which has expired it (so it's a new request for the server), > it could have the potential to cause some kind of corruption then, while at > the same time adding a time component to the key would have prevented the > issue? > > If we really want to protect against this scenario, should a separate > timestamp field be used? This question has been raised by Dennis and Ryan > in the community sync, and Yun above in this thread. > > > > On Wed, Oct 29, 2025 at 7:19 PM yun zou <[email protected]> > wrote: > >> Hi All, >> >> It sounds like the idea is to use the timestamp component to help >> determine whether a key should expire. However, I’m not clear on how >> exactly the timestamp would be used for this purpose. When the server >> decides to expire a key, will it rely on the server-receive timestamp, >> the client request timestamp, or the key creation timestamp? >> >> If the timestamp component within the key is intended to represent the >> key creation time, that raises a couple of concerns: >> 1. A key could be created well before it’s actually used. >> 2. Clock skew between the client and server could lead to inconsistent >> expiration behavior. >> >> If the server is responsible for managing the key lifecycle, it’s >> generally more robust and consistent to rely on the server clock for >> expiration decisions rather than client-provided timestamps. >> >> Additionally, if the timestamp is an important piece of information, >> it might be cleaner to make it an explicit field instead of >> overloading the key itself with multiple purposes. Having a separate, >> well-defined field would make the specification clearer and easier to >> maintain. >> >> From the client’s perspective, requiring the use of UUIDv7 introduces >> unnecessary constraints on implementation. That said, clients are free >> to adopt UUIDv7 if they prefer. Since the server ultimately manages >> expiration, it’s generally better to keep the client logic simple and >> decoupled from server-side decisions. >> >> Best Regards, >> Yun >> >> On Wed, Oct 29, 2025 at 12:50 PM Dmitri Bourlatchkov <[email protected]> >> wrote: >> > >> > Hi All, >> > >> > From my POV (and I may be repeating what I put in GH comments), the >> main point in using UUID v7 is specifying that a timestamp should be part >> of the idempotency key. As previously discussed, having this timestamp is >> beneficial to server implementations. >> > >> > The IETF Idempotency Key draft v7 [1] allows servers to require >> specific ID generation algorithms. >> > >> > We could have a custom ID format, but UUID v7 is already defined and >> fits this use case. >> > >> > If for some reason UUID v7 becomes "weak" in the future, such an event >> will have a much greater impact than the REST Catalog API. In any case, if >> that happens, nothing prevents revisioning the REST API spec to allow for >> stronger ID generators. >> > >> > [1] >> https://datatracker.ietf.org/doc/html/draft-ietf-httpapi-idempotency-key-header-07#name-client >> > >> > Cheers, >> > Dmitri. >> > >> > On Mon, Oct 27, 2025 at 2:33 PM Yufei Gu <[email protected]> wrote: >> >> >> >> +1 on option 2: don’t mandate a specific key format. >> >> >> >> Concerns with option 1 (UUIDv7-mandatory): >> >> 1. Overspecification risk. If UUIDv7 shows weaknesses later, we’re >> stuck with a brittle contract. >> >> 2. Unnecessary constraints. It binds both client and server >> implementations. One of IRC’s goals is to simplify client work; forcing >> UUIDv7 limits client choices for marginal gain (the embedded timestamp). >> >> >> >> Here are existing implementations for reference: >> >> >> >> Stripe[1]: recommends UUIDv4 but does not enforce a format for >> idempotency keys. >> >> AWS EC2[2]: accepts any unique, case-sensitive string up to 64 ASCII >> characters for the client token. >> >> >> >> I'd propose to treat the idempotency key as an opaque string with >> basic requirements and guidance(e.g., “unique string values; UUIDv4 or v7 >> are fine”) but avoid making the format mandatory. This keeps the API >> future-proof and client-friendly while preserving server-side flexibility. >> >> >> >> 1. https://docs.stripe.com/api/expanding_objects >> >> 2. >> https://docs.aws.amazon.com/ec2/latest/devguide/ec2-api-idempotency.html >> >> >> >> Yufei >> >> >> >> >> >> On Mon, Oct 27, 2025 at 9:53 AM huaxin gao <[email protected]> >> wrote: >> >>> >> >>> Hi Yun, >> >>> Thanks for the thoughtful feedback! >> >>> >> >>> Yes, the key itself is expected to be globally unique. You’re also >> right that we don’t need to mandate UUIDs to achieve that; other schemes >> can provide global uniqueness. >> >>> >> >>> I have chosen UUID because several folks in the community prefer it >> as a common, interoperable choice. That said, I agree that mandating UUIDv7 >> adds constraints on clients without clear spec-level benefit. >> >>> >> >>> I also agree we should separate spec from implementation; details >> like the key generation method can live in implementation guidance. >> >>> >> >>> From your note, it sounds like you support Option 2 >> (version-agnostic)—i.e., require a “globally unique idempotency key” and >> accept any RFC 9562 UUID (with v7 as a non-normative recommendation), while >> leaving timestamp/expiry mechanics to the server-side doc. I’ll count this >> as a +1 for Option 2. >> >>> >> >>> Thanks, >> >>> >> >>> Huaxin >> >>> >> >>> >> >>> On Fri, Oct 24, 2025 at 7:00 PM yun zou <[email protected]> >> wrote: >> >>>> >> >>>> Sorry, I accidentally sent the email before complete, please ignore >> my >> >>>> previous email. Sorry for the noise and inconvenience. >> >>>> >> >>>> Hi Huaxin, >> >>>> >> >>>> This is a really interesting and valuable proposal — it provides a >> >>>> great way to address the issue of duplicate client requests. Thank >> you >> >>>> for proposing and driving this forward! >> >>>> >> >>>> One point that isn’t entirely clear to me is how the server uniquely >> >>>> identifies each request. Are we relying solely on the >> idempotency-key >> >>>> being globally unique, or is there an additional identifier such as >> >>>> clientId + idempotency-key? Based on the current discussion, it >> sounds >> >>>> like the proposal expects the key itself to be globally unique, >> likely >> >>>> through the use of a UUID, but I’d like to double-check my >> >>>> understanding. >> >>>> >> >>>> If we are indeed relying on the client to generate a globally unique >> >>>> ID, that approach makes sense. However, it doesn’t seem necessary to >> >>>> mandate the use of UUIDs, as there are other valid methods for >> >>>> achieving global uniqueness. Imposing a further restriction to UUIDv7 >> >>>> would place additional constraints on the client implementation. >> >>>> >> >>>> From a specification perspective, I think it would be better to >> >>>> separate the spec from the implementation. In other words, we should >> >>>> make it clear that the key must be globally unique, but we don’t need >> >>>> to specify that it must be a UUID or UUIDv7. >> >>>> >> >>>> Best Regards, >> >>>> Yun >> >>>> >> >>>> On Fri, Oct 24, 2025 at 4:41 PM huaxin gao <[email protected]> >> wrote: >> >>>> > >> >>>> > Hi all, >> >>>> > >> >>>> > Thank you for taking the time to review my proposal and PR—I >> really appreciate the input. >> >>>> > >> >>>> > There’s one remaining issue I’d like to settle. In the Iceberg >> Catalog Community sync, many preferred mandating UUIDv7 for the idempotency >> key. At the same time, there are some concerns: >> >>>> > >> >>>> > If we need a timestamp, it should be a separate field; we >> shouldn’t use the UUIDv7 timestamp. >> >>>> > >> >>>> > If we use the UUID timestamp for expiry, we’d have to require keys >> to be generated at request time, which feels over-engineered. >> >>>> > >> >>>> > If we want to use the UUIDv7 timestamp, it should be for debugging >> only. >> >>>> > >> >>>> > Based on that, here’s a draft update to the spec: >> >>>> > >> >>>> > Key Requirements: >> >>>> > - Key format: UUIDv7 in string format as defined in RFC 9562. >> >>>> > See >> https://datatracker.ietf.org/doc/html/rfc9562#name-example-of-a-uuidv7-value >> . >> >>>> > - The idempotency key must be globally unique (no reuse across >> different operations). >> >>>> > - Catalogs SHOULD NOT expire keys before the end of the advertised >> token lifetime. >> >>>> > - If Idempotency-Key is used, clients MUST reuse the same key when >> retrying the same >> >>>> > logical operation and MUST generate a new key for a different >> operation. >> >>>> > - Server behavior: Servers MUST validate the syntactic validity of >> UUIDv7 (per RFC 9562). >> >>>> > Servers MUST NOT make behavioral decisions based on the UUID’s >> internal timestamp fields. >> >>>> > The idempotency key is an opaque, unique identifier used only >> for lookup/deduplication. >> >>>> > >> >>>> > This reads a bit awkward to me: we mandate UUIDv7 but prohibit >> using its timestamp, which seems to undercut the reason to require v7 in >> the first place. >> >>>> > >> >>>> > I’d appreciate feedback on whether we should: >> >>>> > >> >>>> > Option 1 — Require v7. >> >>>> > Keep UUIDv7 required, with the server restrictions above >> (syntactic v7 validation only; no behavioral decisions based on the >> embedded timestamp). >> >>>> > >> >>>> > Option 2 — Version-agnostic. >> >>>> > Make the client spec version-agnostic (require RFC 9562 UUID >> textual form; allow v7 as a recommendation). Leave any timestamp/lifetime >> mechanics to a server-side (Polaris idempotency) document. >> >>>> > >> >>>> > Thanks again for the thoughtful discussion. >> >>>> > >> >>>> > Best, >> >>>> > >> >>>> > Huaxin >> >>>> > >> >>>> > >> >>>> > On Mon, Sep 29, 2025 at 5:47 PM Dmitri Bourlatchkov < >> [email protected]> wrote: >> >>>> >> >> >>>> >> Hi Huaxin, >> >>>> >> >> >>>> >> Sorry about the delay. I posted some comments on >> https://github.com/apache/iceberg/pull/14196 Some of them I might have >> mentioned on the doc too, so apologies if they got answered in the doc and >> I missed it. >> >>>> >> >> >>>> >> Cheers, >> >>>> >> Dmitri. >> >>>> >> >> >>>> >> On Thu, Sep 25, 2025 at 12:27 PM huaxin gao < >> [email protected]> wrote: >> >>>> >>> >> >>>> >>> Thank you all for taking the time to review and discuss! I’ve >> responded to all questions and updated the proposal. If there are no >> additional concerns, I’ll proceed to start a VOTE thread. >> >>>> >>> >> >>>> >>> Thanks, >> >>>> >>> Huaxin >> >>>> >>> >> >>>> >>> On Mon, Sep 22, 2025 at 1:30 AM Maninder Parmar < >> [email protected]> wrote: >> >>>> >>>> >> >>>> >>>> +1, for low level retry which ensures that the idempotent key >> is never committed twice. I also agree that canonicalizing the request body >> where the client can change it due to conflict resolution and retry would >> be hard to get right. >> >>>> >>>> >> >>>> >>>> On Sat, Sep 20, 2025 at 5:58 AM Dennis Huo <[email protected]> >> wrote: >> >>>> >>>>> >> >>>> >>>>> +1 to this being mostly targeting a "low-level" retry >> semantic. Expanding on that though I'd say even "client-side retries" >> really have two distinct flavors: >> >>>> >>>>> >> >>>> >>>>> A. Business-logic-agnostic retries, e.g. in a common low-level >> HTTP client library - behaviorally, these should behave largely the same as >> "network infra retries". The key distinction is that in this case any >> content hashing would be *post* serialization and even agnostic to >> request-body content-type (i.e. not JSON-specific). >> >>>> >>>>> B. Application-specific retries, such as when Iceberg client >> will potentially rebase on a new snapshot >> >>>> >>>>> >> >>>> >>>>> I think this aligns with what Peter and others mentioned >> earlier where trying to canonicalize the *semantic* content of a request is >> probably brittle/risky. And as Yufei mentions, case 2.B (client-side real >> application-layer retries) should be using a new idempotency-key if it's >> ever doing the retry at the later that requires re-serializating JSON. >> >>>> >>>>> >> >>>> >>>>> Overall though I agree making the content-hash checking >> optional is a good idea. >> >>>> >>>>> >> >>>> >>>>> On Fri, Sep 19, 2025 at 4:33 PM huaxin gao < >> [email protected]> wrote: >> >>>> >>>>>> >> >>>> >>>>>> Thanks, Peter and Yufei. I agree the main use case is >> network‑infrastructure retries. To keep the specification simple and move >> the proposal forward, let’s make the baseline key‑only idempotency. If >> there’s demand, we can add an optional payload‑binding mode (canonical JSON >> + SHA‑256), advertised via /v1/config. >> >>>> >>>>>> >> >>>> >>>>>> Thanks, >> >>>> >>>>>> >> >>>> >>>>>> Huaxin >> >>>> >>>>>> >> >>>> >>>>>> >> >>>> >>>>>> On Fri, Sep 19, 2025 at 1:31 PM Yufei Gu < >> [email protected]> wrote: >> >>>> >>>>>>> >> >>>> >>>>>>> "Network infrastructure retries" would be the dominant use >> case. I'd NOT recommend clients retry with the same idempotency key if it >> regenerated the request, instead, clients should reload before retry in >> that case. >> >>>> >>>>>>> >> >>>> >>>>>>> Yufei >> >>>> >>>>>>> >> >>>> >>>>>>> >> >>>> >>>>>>> On Fri, Sep 19, 2025 at 2:05 AM Péter Váry < >> [email protected]> wrote: >> >>>> >>>>>>>> >> >>>> >>>>>>>> Hi Huaxin, >> >>>> >>>>>>>> >> >>>> >>>>>>>> Could you clarify the specific use cases we intend to >> support regarding retry checking? Here are a couple of possibilities I had >> in mind: >> >>>> >>>>>>>> >> >>>> >>>>>>>> Network infrastructure retries – where the exact same >> request is retried. >> >>>> >>>>>>>> Client-side retries – where the client regenerates the >> request using the same program logic, resulting in identical content. >> >>>> >>>>>>>> >> >>>> >>>>>>>> If there are no security or other concerns, I’d suggest >> keeping the specification simple and avoiding mechanisms that surface >> client-side implementation errors. The cleanest approach might be to ignore >> the request content and rely solely on a user-provided key. >> >>>> >>>>>>>> >> >>>> >>>>>>>> Alternatively, we could include an optional error code in >> the response, which implementations may use to signal conflicts. The actual >> conflict detection logic can be left to the implementations—we don’t need >> to define it in the specification. If we go this route, we should also >> offer a way to disable these checks, since there will inevitably be cases >> where semantically identical requests are incorrectly flagged as >> conflicting. >> >>>> >>>>>>>> >> >>>> >>>>>>>> Thanks, >> >>>> >>>>>>>> Peter >> >>>> >>>>>>>> >> >>>> >>>>>>>> huaxin gao <[email protected]> ezt írta (időpont: >> 2025. szept. 19., P, 1:38): >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Thanks Steven for the +1 and for raising the fingerprint >> question! Great points! >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> What we need to protect against: >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Same logical request, different bytes across retries >> (pretty vs compact JSON, map key order, ...). >> >>>> >>>>>>>>> Accidental key reuse with a changed payload. >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Options and tradeoffs: >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Exact byte checksum (e.g., SHA‑256 over raw body) >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Pro: trivial, fast >> >>>> >>>>>>>>> Con: too strict; benign diffs cause false mismatches >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Canonical JSON over full request, then hash (proposed) >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Pro: stable across whitespace/key order; simple to >> implement for typed payloads >> >>>> >>>>>>>>> Con: slightly more work than raw checksum; >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Checksum of selected fields / field-by-field match >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Pro: can be faster for huge payloads; can ignore noisy >> fields >> >>>> >>>>>>>>> Con: could misses legitimate differences >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Request digest/signature >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Pro: very strong >> >>>> >>>>>>>>> Con: heavyweight >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Maybe we could make this configurable: >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> canonical-json-sha256 (default) >> >>>> >>>>>>>>> raw-bytes-sha256 (strict) >> >>>> >>>>>>>>> trust-client-key (no fingerprint check) >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> On the IETF draft status: >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> I have also noted the draft’s expiry. We will align with >> its semantics for now and can adjust if a new version lands. >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Thanks, >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> Huaxin >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> >> >>>> >>>>>>>>> On Thu, Sep 18, 2025 at 4:01 PM Steven Wu < >> [email protected]> wrote: >> >>>> >>>>>>>>>> >> >>>> >>>>>>>>>> +1 for the feature that can make retry safe for 500s and >> improve the client fault-tolerance of transient server failures. >> >>>> >>>>>>>>>> >> >>>> >>>>>>>>>> Peter and Dimitri raised a good question on the >> fingerprint. The IETF draft doesn't actually define the fingerprint algo. >> We can also go with simple checksum of the entire request payload, which >> would be cheap to compute. Do we anticipate any anticipated scenarios where >> clients may rewrite the payload in different forms of serialized bytes >> during retries? >> >>>> >>>>>>>>>> >> >>>> >>>>>>>>>> * Checksum of the entire request payload. >> >>>> >>>>>>>>>> * Checksum of selected element(s) in the request >> payload. >> >>>> >>>>>>>>>> * Field value match for each field in the request >> payload. >> >>>> >>>>>>>>>> * Field value match for selected element(s) in the >> request payload. >> >>>> >>>>>>>>>> * Request digest/signature >> >>>> >>>>>>>>>> >> >>>> >>>>>>>>>> >> >>>> >>>>>>>>>> BTW, the IETF draft seems to have expired without approval >> >>>> >>>>>>>>>> >> https://datatracker.ietf.org/doc/draft-ietf-httpapi-idempotency-key-header/ >> >>>> >>>>>>>>>> >> >>>> >>>>>>>>>> On Thu, Sep 18, 2025 at 3:46 PM huaxin gao < >> [email protected]> wrote: >> >>>> >>>>>>>>>>> >> >>>> >>>>>>>>>>> Thanks Peter and Dmitri for the thoughtful feedback! I >> really appreciate you taking a close look at my proposal. I agree that >> "semantic equality" is tricky, that's why the scope here is intentionally >> narrow. >> >>>> >>>>>>>>>>> >> >>>> >>>>>>>>>>> Just to clarify scope: I’m not trying to solve general >> semantic equivalence. For these specific, typed request payloads, I >> serialize to a deterministic JSON and hash it. That normalizes benign diffs >> (map order, whitespace) without trying to infer meaning. The goal is a >> stable fingerprint so that if a key is accidentally reused with a changed >> payload, we surface that instead of silently diverging. >> >>>> >>>>>>>>>>> >> >>>> >>>>>>>>>>> To make this feel less brittle, I’ll add tests for the >> practical cases (ordering/whitespace, nested maps, a clear null‑vs‑missing >> rule, numeric formatting), plus end‑to‑end tests in the in‑memory REST >> fixture with failure injection (in‑flight dup, finalize failure -> >> reconcile, etc.). Happy to walk through these if helpful. >> >>>> >>>>>>>>>>> >> >>>> >>>>>>>>>>> I’m also open to adding a config switch for >> “trust‑client‑key only” if that’s preferred in some environments. My intent >> is to stay aligned with the IETF Idempotency‑Key guidance (first request >> wins; conflicting reuse is rejected, and reusing a key with a different >> request payload is rejected via an idempotency fingerprint) while keeping >> things as simple as possible and protecting us from accidental key misuse. >> Would love to align on the lightest approach that meets those goals. >> >>>> >>>>>>>>>>> >> >>>> >>>>>>>>>>> Thanks, >> >>>> >>>>>>>>>>> >> >>>> >>>>>>>>>>> Huaxin >> >>>> >>>>>>>>>>> >> >>>> >>>>>>>>>>> >> >>>> >>>>>>>>>>> On Thu, Sep 18, 2025 at 6:17 AM Dmitri Bourlatchkov < >> [email protected]> wrote: >> >>>> >>>>>>>>>>>> >> >>>> >>>>>>>>>>>> Hi All, >> >>>> >>>>>>>>>>>> >> >>>> >>>>>>>>>>>> I agree that checking request contents is almost >> redundant in this case. >> >>>> >>>>>>>>>>>> >> >>>> >>>>>>>>>>>> If the randomness quality of Idempotency-Key value is >> good, collisions are very unlikely on the server side. Given that, any >> content checks the server performs are essentially validating that clients >> correctly reuse the generated Idempotency-Key value. (this is mostly the >> same as my comment on the related Polaris discussion). >> >>>> >>>>>>>>>>>> >> >>>> >>>>>>>>>>>> I'd like to propose making the content check optional >> so that servers may or may not implement it according to their design >> principles and constraints and emphasizing that clients should use unique >> keys (e.g. UUIDs)... basically going with option 2 from Peter's email. >> >>>> >>>>>>>>>>>> >> >>>> >>>>>>>>>>>> I believe this is in line with the SHOULD word used for >> this case in the IETF draft [1] (section 2.7). >> >>>> >>>>>>>>>>>> >> >>>> >>>>>>>>>>>> [1] >> https://datatracker.ietf.org/doc/html/draft-ietf-httpapi-idempotency-key-header-06 >> >>>> >>>>>>>>>>>> >> >>>> >>>>>>>>>>>> Thanks, >> >>>> >>>>>>>>>>>> Dmitri. >> >>>> >>>>>>>>>>>> >> >>>> >>>>>>>>>>>> On Thu, Sep 18, 2025 at 7:56 AM Péter Váry < >> [email protected]> wrote: >> >>>> >>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>> Thanks Huaxin for the proposal, and sorry for the late >> review - I had a bit of a busy week. >> >>>> >>>>>>>>>>>>> I have one main question, which I have also added as a >> comment to the doc: >> >>>> >>>>>>>>>>>>> - Why do we try to compare the request contents when >> the Idempotency-Key is the same for the requests? The comparison algorithm >> is a bit complicated, and seems brittle to me. Consistent field ordering, >> maps, and maybe even inconsistency in upper case/lower case letters might >> mean technically the same request. >> >>>> >>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>> In my previous roles (admittedly more than 10 years >> ago) I was extensively working on APIs like this, and we have never really >> succeeded in creating a good enough "are these 2 requests are really the >> same semantically" checks. >> >>>> >>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>> I would simplify these requirements, unless there are >> serious arguments for the existence of these checks: >> >>>> >>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>> Either check for exact matches - without any magic - >> this could be used for detecting issues where the duplication happens on >> the network side, or >> >>>> >>>>>>>>>>>>> Rely entirely on the clients to provide the correct >> Idempotency-Key. >> >>>> >>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>> I would prefer the 2nd. >> >>>> >>>>>>>>>>>>> Otherwise I agree with the contents of the proposal. >> It is nicely done! (edited) >> >>>> >>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>> Yufei Gu <[email protected]> ezt írta (időpont: >> 2025. szept. 18., Cs, 2:54): >> >>>> >>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>> Thanks for the proposal. It's a nice feature to make >> retry more reliable and efficient. Left some comments. >> >>>> >>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>> Yufei >> >>>> >>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>> On Mon, Sep 15, 2025 at 3:53 PM Kevin Liu < >> [email protected]> wrote: >> >>>> >>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>> Thanks for writing up the proposal! Makes sense to >> add idempotency to mutation requests. >> >>>> >>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>> It would be helpful to add this feature to both the >> catalog test framework and the iceberg-rest-fixture. The latter is used by >> the subprojects for testing and would come in handy when we want to test >> out the client implementation. >> >>>> >>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>> For other reviewers, the Stripe documentation on >> idempotency was a helpful read, >> https://docs.stripe.com/api/idempotent_requests. >> >>>> >>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>> Best, >> >>>> >>>>>>>>>>>>>>> Kevin Liu >> >>>> >>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>> On Mon, Sep 15, 2025 at 11:38 AM Szehon Ho < >> [email protected]> wrote: >> >>>> >>>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>>> Hi, >> >>>> >>>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>>> Sounds like fairly standard practice and makes >> sense to me in the first read. >> >>>> >>>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>>> Thanks, >> >>>> >>>>>>>>>>>>>>>> Szehon >> >>>> >>>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>>> On Mon, Sep 15, 2025 at 10:09 AM Russell Spitzer < >> [email protected]> wrote: >> >>>> >>>>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>>>> I think based on the feedback on the proposal and >> in recent syncs we should probably move forward with the actual Spec Change >> PR so we can see what this looks like and move on to a discussion of how >> the Catalog test framework should test this. >> >>>> >>>>>>>>>>>>>>>>> >> >>>> >>>>>>>>>>>>>>>>> On 2025/08/22 18:26:23 huaxin gao wrote: >> >>>> >>>>>>>>>>>>>>>>> > Hi all, >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > I’d like to propose a change to Iceberg’s REST >> API to make mutation >> >>>> >>>>>>>>>>>>>>>>> > requests safely retryable. >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > *The Problem* >> >>>> >>>>>>>>>>>>>>>>> > If a POST mutation (e.g., updateTable) succeeds >> in the catalog but the >> >>>> >>>>>>>>>>>>>>>>> > client doesn’t receive the response (timeout, >> connection closed, etc.), a >> >>>> >>>>>>>>>>>>>>>>> > second attempt can hit 409 Conflict. The client >> interprets the 409 as a >> >>>> >>>>>>>>>>>>>>>>> > failed commit and deletes the associated >> metadata files, causing >> >>>> >>>>>>>>>>>>>>>>> > catalog/storage inconsistency. >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > *The Proposed Solution* >> >>>> >>>>>>>>>>>>>>>>> > Introduces an optional Idempotency-Key HTTP >> header on REST mutation >> >>>> >>>>>>>>>>>>>>>>> > endpoints and has the Iceberg client pass it >> through. >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > *Semantics *(first processed request wins): >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > - >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > Same key + same canonical payload -> return >> the original result (no >> >>>> >>>>>>>>>>>>>>>>> > re-execution). >> >>>> >>>>>>>>>>>>>>>>> > - >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > Same key + different payload -> 422 >> (Unprocessable Content). >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > *Capability discovery:* catalogs can advertise >> support and retention so >> >>>> >>>>>>>>>>>>>>>>> > clients know when a retry is safe, e.g. >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > { >> >>>> >>>>>>>>>>>>>>>>> > "idempotency-tokens-respected": true, >> >>>> >>>>>>>>>>>>>>>>> > "idempotency-token-lifetime": "30m" } >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > *Scope in Iceberg:* update the OpenAPI to >> include the header, and add >> >>>> >>>>>>>>>>>>>>>>> > client pass-through + honoring capability >> discovery. No server >> >>>> >>>>>>>>>>>>>>>>> > implementation is mandated—catalogs (e.g., >> Polaris) can implement >> >>>> >>>>>>>>>>>>>>>>> > storage/TTL/replay as they choose. >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > *Standards alignment:* uses the >> industry-standard header name and matches >> >>>> >>>>>>>>>>>>>>>>> > the IETF HTTPAPI Idempotency-Key draft >> >>>> >>>>>>>>>>>>>>>>> > < >> https://datatracker.ietf.org/doc/html/draft-ietf-httpapi-idempotency-key-header >> > >> >>>> >>>>>>>>>>>>>>>>> > semantics. >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > *Compatibility:* fully backward compatible. >> Servers that don’t support it >> >>>> >>>>>>>>>>>>>>>>> > can ignore the header; clients can detect >> support via capability discovery. >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > Here is the proposal >> >>>> >>>>>>>>>>>>>>>>> > < >> https://docs.google.com/document/d/1WyiIk08JRe8AjWh63txIP4i2xcIUHYQWFrF_1CCS3uw/edit?tab=t.0 >> >. >> >>>> >>>>>>>>>>>>>>>>> > Looking forward to your thoughts. >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > Thanks, >> >>>> >>>>>>>>>>>>>>>>> > >> >>>> >>>>>>>>>>>>>>>>> > Huaxin >> >>>> >>>>>>>>>>>>>>>>> > >> >
