Hi Let’s Encrypt Community,
I recently built Merkle Lens (https://merklens.app/), a small browser-based tool to help me learn about and explore Certificate Transparency logs. While testing it, I found that I couldn’t read Let’s Encrypt’s Sycamore and Willow logs directly from the browser because the monitoring responses I tested did not include an Access-Control-Allow-Origin header.
Would you consider enabling CORS on these public read endpoints? I’ve included my findings and a couple of questions below.
What I measured
On October 10, 2026, I tested read endpoints for every live log in the Chrome log list. I sent each request with an Origin: https://merklens.app header and recorded the HTTP status and the presence and value of Access-Control-Allow-Origin.
The table below shows the results for Let’s Encrypt’s logs, with one Geomys log included for comparison. Each column represents a static-ct-api endpoint: the checkpoint, a level-0 hash tile, a data tile, and an issuer request using a placeholder fingerprint.
Each cell shows the HTTP status followed by the CORS header value. * means any origin is allowed; none means the header was absent.
| Log | checkpoint |
tile/0/000 |
tile/data/000 |
issuer/<fingerprint> |
|---|---|---|---|---|
| Sycamore2026h2 | 200, none | 200, none | 200, none | 403, none |
| Sycamore2027h1 | 200, none | 200, none | 200, none | 403, none |
| Sycamore2027h2 | 200, none | 200, none | 200, none | 403, none |
| Willow2026h2 | 200, none | 200, none | 200, none | 403, none |
| Willow2027h1 | 200, none | 200, none | 200, none | 403, none |
| Willow2027h2 | 200, none | 200, none | 200, none | 403, none |
| Geomys Tuscolo2026h2 | 200, * |
200, * |
200, * |
404, * |
For the Let’s Encrypt logs, the checkpoint and tile requests succeeded, but the missing CORS header prevents browser JavaScript from reading their responses. The issuer requests returned 403. Since those requests used a placeholder fingerprint, they do not establish how valid issuer requests would behave.
For comparison, the responses I tested from Geomys’s Tuscolo and Trastevere logs, as well as IPng Networks and Microsec, included Access-Control-Allow-Origin: *. I believe Tuscolo and Trastevere also run Sunlight, so I wondered whether this might be a serving or edge-configuration difference rather than a limitation of the log software.
Across the full survey, 38 of the 66 live logs included Access-Control-Allow-Origin on the read endpoints I tested.
The request
Would you consider adding Access-Control-Allow-Origin: * to at least the monitoring endpoints (checkpoint, tile/*, and issuer/*), so browser-based clients can read the public log data directly? Enabling CORS on the submission endpoints as well would be a welcome addition, although my main request is for monitoring access.
Access-Control-Allow-Origin: *, so browser-based clients can read the public log data directly.Access-Control-Expose-Headers: Retry-After, so those clients can readRetry-After, when provided, and respect the requested retry interval.
I did not find Retry-After exposed through Access-Control-Expose-Headers in any of the responses I tested.
The goal is simply to let browser-based tools access the same public data that other HTTP clients can already retrieve, without needing an intermediary proxy.
Two questions for the wider community
I would also appreciate some guidance on raising this more broadly:
- Where would be the best place to discuss CORS support across log operators? Besides contacting operators individually using the addresses in the Chrome log list, is there a shared venue they are likely to follow? I was considering posting the survey to
ct-policy@chromium.org, but not sure if that be appropriate. - What is the preferred way to propose a change to a C2SP specification? The Static Certificate Transparency API specification includes
SHOULDrecommendations for caching headers but does not mention CORS. I’d like to propose a similar recommendation forAccess-Control-Allow-Origin: *on monitoring endpoints. Should I start with an issue or pull request in the C2SP repository, or discuss it with the maintainers first?
I’m happy to share the survey script and the full per-endpoint results for all operators. I’ll also re-test and report back here if anything changes.
Thanks for operating these logs, and for any advice you can offer!