- From: Rory Hewitt <rory.hewitt@gmail.com>
- Date: Mon, 20 Jul 2026 11:56:00 -0700
- To: Joost de Valk <joost@joost.blog>
- Cc: ietf-http-wg@w3.org
- Message-ID: <CAEmMwDzvaMutoej+ZYV+ZiZcVEBaVHRFwLsWxFGVsAio6UBh+g@mail.gmail.com>
To be very clear, I don't think you need any changes to your document - it's more about how this is eventually documented to end-users - when should it be added, when should it be overwritten, and when should it be *trusted* (never). From a technical perspective, your doc is fine. IMO, obviously. On Mon, Jul 20, 2026 at 11:24 AM Joost de Valk <joost@joost.blog> wrote: > I would fully agree with that - happy to make that change. > > On Mon, Jul 20, 2026 at 8:21 PM Rory Hewitt <rory.hewitt@gmail.com> wrote: > >> I like this. >> >> To be clear, there's no requirement for any of those intermediaries to >> register 'their' value? For example, it may be that a CDN called "The Cool >> CDN" might always add `Redirect-By: coolcdn`or `Redirect-By: cool` or >> `Redirect-By: cdn` or even `Redirect-By: "Blah blah blah"` - they can >> choose to use whatever value they want, although it would be in everyone's >> best interests for companies to standardize on a value they choose to >> specify? >> >> Obviously this means that if a hacker were able to MitM the response >> processing, they could add the `Redirected-By: coolcdn` response header, >> and hope that the client would incorrectly believe that the redirect was >> 'legitimate'? >> >> I think the only concern would be making it very very very clear that >> this header can't be trusted, and is really most useful when debugging >> request problems. >> >> On Mon, Jul 20, 2026 at 7:07 AM Joost de Valk <joost@joost.blog> wrote: >> >>> Hi all, >>> >>> First time e-mailer, but I thought I'd give it a try :) >>> >>> I've submitted a first draft that defines the Redirect-By response >>> header field: https://datatracker.ietf.org/doc/draft-devalk-redirect-by/ >>> >>> Redirect-By lets the component whose decision determined a redirect >>> identify itself in the response. When a redirect is misconfigured, the >>> status code and Location tell you nothing about which of the CDN rules, >>> reverse proxies, CMS logic, or plugins in the chain produced it, so >>> diagnosis today usually means disabling components one at a time. >>> >>> This field turns an opaque redirect into a self-describing one. This >>> standardizes existing practice. The convention originated during The >>> Guardian's migration from guardian.co.uk to theguardian.com (now quite >>> a while ago) and has shipped as X-Redirect-By in WordPress core since >>> version 5.1 (sent by default on every redirect, with a filter so plugins >>> can identify themselves), in TYPO3's redirect middleware, and in a growing >>> list of other implementations. >>> >>> Two design choices I'd particularly welcome feedback on: >>> >>> 1. The draft doesn't just add an unprefixed alias, it defines >>> Redirect-By as a Structured Field (an RFC 9651 Item, Token or String) and >>> formally deprecates X-Redirect-By, requesting registration of the legacy >>> name with status "deprecated". The reasoning: migration already requires >>> senders to adopt a new name, so adopting the standard serialization at the >>> same time costs little, and it gives well-defined parsing plus an >>> unambiguous rule for duplicate or combined field lines. >>> >>> 2. The field identifies the component that determined the Location field >>> specifically. An intermediary that rewrites only the status code leaves the >>> field alone; one that changes the Location must replace or remove it. >>> >>> HTML rendering and issue tracker: >>> >>> - https://jdevalk.github.io/draft-devalk-redirect-by/ >>> - https://github.com/jdevalk/draft-devalk-redirect-by/issues >>> >>> Comments here or as issues are both very welcome! >>> >>> Joost de Valk >>> >> >> >> -- >> Rory Hewitt >> >> https://www.linkedin.com/in/roryhewitt >> > -- Rory Hewitt https://www.linkedin.com/in/roryhewitt
Received on Monday, 20 July 2026 18:56:17 UTC