Hi Eric,

Thank you, I'll post there too!

Best,
Pierre

On Sun, Jun 29, 2025, at 19:08, Eric Rescorla wrote:
> Hi Pierre,
> 
> This is pretty far afield from TLS. I would suggest the secdispatch mailing 
> list.
> 
> Best,
> -Ekr
> 
> 
> On Sun, Jun 29, 2025 at 3:45 AM Pierre Barre <[email protected]> wrote:
>> __
>> Hi all,
>> 
>> I've just submitted a new individual draft that extends RFC6962 Certificate 
>> Transparency with a page-based retrieval mechanism:
>> 
>> https://datatracker.ietf.org/doc/html/draft-trans-pages-00
>> 
>> This proposal offers a simpler alternative to the "Static API" approach 
>> while maintaining full RFC 6962 compatibility. Unlike the Static API which 
>> requires significant client changes and complexity, this is just an 
>> incremental extension that adds optional endpoints to existing CT logs.
>> 
>> The draft addresses several operational challenges with the current CT 
>> get-entries API:
>> 
>> 1. Arbitrary range requests make caching effectively impossible
>> 2. Base64 encoding adds ~33% bandwidth overhead  
>> 3. Certificate chains are duplicated for every entry
>> 4. Variable response sizes complicate resource planning
>> 
>> The extension introduces:
>> - Fixed-size pages (e.g., 1000 entries) that are immutable once full
>> - Binary format eliminating base64 overhead
>> - Certificate deduplication via SHA-256 references
>> - Simple discovery mechanism
>> - Full backward compatibility - existing RFC 6962 clients continue to work
>> 
>> Key advantages over Static API:
>> - Much simpler for clients to implement
>> - No breaking changes to existing CT infrastructure
>> - Allows incremental adoption
>> - Enables CDN deployment while keeping the familiar CT log model
>> 
>> I'm particularly interested in feedback from:
>> - CT log operators on operational benefits
>> - Client implementers on the API design
>> - Anyone with Static API implementation experience for comparison
>> 
>> I've implemented this in my CT log: https://github.com/Barre/compact_log
>> 
>> 
>> Performance comparison on a real log with 279,332 entries shows the Pages 
>> Extension is more efficient than Static CT API:
>> 
>> Pages Extension:
>> - 280 requests (vs 1,092 for Static CT)  
>> - 212 MB transfer (vs 243 MB for Static CT)
>> - 756 bytes/entry (vs 869 bytes/entry)
>> - Sequential read test: 109 seconds (vs 143 seconds)
>> - Uses only 87.2% of Static CT bandwidth
>> 
>> The binary format also compresses well with standard HTTP compression:
>> - Brotli: 56% size reduction
>> - Gzip: 58% size reduction
>> 
>> Since the trans WG has concluded, I'm bringing this to TLS as the most 
>> relevant venue for CT implementation discussions.
>> 
>> Thanks,
>> Pierre
>> _______________________________________________
>> TLS mailing list -- [email protected]
>> To unsubscribe send an email to [email protected]
_______________________________________________
TLS mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to