| Internet-Draft | dereferenceable identifiers | March 2024 |
| Bormann & Amsüss | Expires 3 September 2024 | [Page] |
In a protocol or an application environment, it is often important to be able to create unambiguous identifiers for some meaning (concept or some entity).¶
Due to the simplicity of creating URIs, these have become popular for this purpose. Beyond the purpose of identifiers to be uniquely associated with a meaning, some of these URIs are in principle dereferenceable, so something can be placed that can be retrieved when encountering such a URI.¶
The present -03 revision discusses the continuum between entirely relying on the result of dereferencing an identifier and building in knowledge of all expected identifiers, and it mentions further pitfalls of using dereferenceable identifiers.¶
This note is to be removed before publishing as an RFC.¶
Status information for this document may be found at https://datatracker.ietf.org/doc/draft-bormann-t2trg-deref-id/.¶
Discussion of this document takes place on the t2trg Research Group mailing list (mailto:t2trg@irtf.org), which is archived at https://mailarchive.ietf.org/arch/browse/t2trg/. Subscribe at https://www.ietf.org/mailman/listinfo/t2trg/.¶
Source for this draft and an issue tracker can be found at https://github.com/cabo/deref-id.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 3 September 2024.¶
Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document.¶
(Please see abstract.)¶
Identifier¶
Dereferenceable¶
Dereference¶
The information that is retrieved by dereferencing a dereferenceable identifier.¶
Dereferencing requires some server infrastructure to actually provide the information. Simplifying the potential complexity of this infrastructure, the entity (entities) controlling the operation of this server infrastructure, including the name spaces in use (e.g., DNS names, URI paths on a server) are called the operator(s) of the dereferenceable identifier.¶
An entity that receives data containing a dereferenceable identifier.¶
A directed identifier is an identifier that has been specifically minted to not just identify the intended entity, but also context information such as the intended use, or intended consumer of the identifier.¶
Directed information is information that is tailored to the implicit context of a specific dereferencing access, such as the accessing IP address or other ancillary parameters. (Content negotiation alone is not "directed information", as it is explicitly triggered by the dereferencing entity.)¶
A unique identifier is an identifier that is unique for the entity; i.e., no other identifiers are in use (or intended to be in use).¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [BCP14] when, and only when, they appear in all capitals, as shown here.¶
This section is intended to present a number of examples where dereferenceable identifiers are in use in a protocol, including existing discussion about constraints on their usage, the benefits claimed for this constrained usage, and remaining issues.¶
Many protocols based on XML or JSON include a protocol or protocol version identifier in the heading to a data item.¶
E.g., [JSO] defines a language for data models that contain an
identifier to the language version in use, here
https://json-schema.org/draft/2020-12/schema.
The model that can be retrieved from this URI in turn contains
further dereferenceable identifiers that point to further details.¶
Section 8.1.1 of [JSO] has this:¶
If this URI identifies a retrievable resource, that resource SHOULD be of media type "application/schema+json".¶
So it acknowledges that the dereferenceability is optional, but does place further restrictions on what can be the result of a successful dereference: another one of these data models, which in turn contain further dereferenceable identifiers.¶
The Problem Details for HTTP APIs format [PROBLEM] uses a dereferenceable identifier for its "type" field. The value is a URI that "identifies the specific "problem type" (e.g., "out of credit")" (Section 1 of [PROBLEM]).¶
Section 3.1.1 of [PROBLEM] has this:¶
If the type URI is a locator (e.g., those with an "http" or "https" scheme), dereferencing it SHOULD provide human-readable documentation for the problem type (e.g., using HTML [HTML5]).¶
but then warns:¶
However, consumers SHOULD NOT automatically dereference the type URI, unless they do so when providing information to developers (e.g., when a debugging tool is in use).¶
Section 4 of [PROBLEM] further details:¶
A problem type URI SHOULD resolve to HTML [HTML5] documentation that explains how to resolve the problem.¶
This becomes even more interesting as Section 4.2 of [PROBLEM] then gives this advice:¶
Registrations MAY use the prefix "https://iana.org/assignments/http-problem-types#" for the type URI.¶
A reference to the place where registrations for these items are managed is certainly desirable, however, the implications on the management of fragment identifiers in the HTML documents that IANA generates from registration information are an example for the increased complexity dereferenceable identifiers may place on the owners of the URI space employed.¶
There are a lot more examples in published RFCs; add them to this document.¶
If a data item containing dereferenceable identifier(s) becomes widely distributed, naive implementations that handle such a data item might dereference these identifiers as part of a routine operation. Many definitions of dereferenceable identifiers contain admonitions that such a behavior can cause an implosion of requests on the server(s) for the URI.¶
Dereferenceable URIs usually contain domain names, whose ownership can change. As a result, and for other reasons as well, parts of the name space of an origin may come under new administration, which can change the policies that apply to resources made available there.¶
These are problems of such URIs in general (and can be mitigated by going to a non-dereferenceable kind of URIs such as one based on the 'tag' uri scheme [TAG]). However, the problems are exacerbated by their use as a dereferenceable identifier. The new owner/administrator might more easily accept that a certain chunk of their URI space should not be used (which suffices for a non-dereferenceable identifier based on this kind of URI namespace) than that certain content needs to be offered there (potentially presenting non-trivial loads, some mechanisms needed to update that information, and legal liabilities that are hard to assess).¶
Dereferencing an identifier may produce different representations over time. While these changes may be intentional and beneficial (e.g. because terms are compatibly added to a resource describing terms that are evolved together [COOL]), they can also cause breakage in applications that previously dereferenced the identifier successfully:¶
There can be errors in the representation introduced by the change.¶
The operator and the consumer may disagree about what constitutes a compatible change.¶
An updated representation may exceed the consumer's capabilities, e.g. not fitting in an allocated buffer.¶
Even without intended changes to the representation, changes to the channel may exclude certain consumers. For example, the operator's web server may cease to accept the cipher suites implemented in the consumer.¶
When the operator's services are compromised, there may be malicious changes in the representation.¶
Dereferencing an identifier may involve following some redirections; whether that following is actually implied, or desired (or even desirable) is rarely being discussed.¶
Denial of service attacks are discussed in Section 6. Privacy implications, in particular around single-use identifiers, are discussed in Section 7.¶
Consumers do not face a binary choice between dereferencing dereferenceable identifiers and treating them as opaque. The space between those extremes is continuous. Notable steps consumers can take to mitigate pitfalls of dereferencing are:¶
Consumers that dereference may apply caching, which reduces server load and bridges both outages and misconfigurations on the server side.¶
These caches may adhere to the caching rules of the underlying systems (DNS result life times, HTTP's freshness rules), but may also stretch them if the alternative are failures or treating the identifier as opaque.¶
Consumers may use caching proxy services provided by trusted parties.¶
While this increases the susceptibility to service outages, it immediately mitigates the privacy implications of having the consumer's network address visible to the operator. Restrictive policies at the proxy can further mitigate other issues. For example, if the proxy's cache is eagerly populated by web spider operations from public starting points and only ever serves cached results to consumers, it defends against single-use URIs.¶
Consumer caches may be pre-populated as part of their firmware update mechanisms.¶
In its extreme form, the consumer may not even be equipped to dereference any identifiers outside of its cache, and the dereferenced representation may already be part of the firmware in ingested form to save runtime resources. Such a consumer shares its properties with a consumer that treats dereferenceable identifiers as opaque. However, the authors of the firmware can make good use of the dereferenceable identifiers. For example, they can dereference a known (or spidered) set of identifiers in an automated fashion, with any suitable amount of caching or manual verification.¶
This document makes no concrete requests on IANA, but does point out that IANA resources might be a good target for a certain class of dereferenceable identifiers.¶
The ability to create a denial of service attack by pointing a dereferenceable identifier into a popular data item that is widely distributed is implied by the discussion in Section 2, alongside with some recommendations for implementers that would mitigate such attacks. A problem with such recommendations is that they need to be followed by implementations that are using dereferenceable identifiers, which might not care much.¶
Dereferencing an identifier leaves a wide-spread data trail, ranging from host name lookups visible on the network to the absolute URI (i.e., the URI without its fragment identifier) visible to the operator of the identifier. Moreover, the operator might gain additional data about the requester, e.g. from a User-Agent header.¶
By minting directed (e.g., single-use) dereferenceable identifiers and assigning short cache lifetimes to the dereferenced resource, the originator of a document can track dereferencing clients whenever they process the document the identifier has been created for. Moreover, single-use identifiers can also be used to exfiltrate data from originators whose network access is restricted through dereferencing clients.¶
Christian Amsüss pointed out that this document would be good to have.¶