atproto aha moments:
- you can host your data
- it's just json
- it's typed but any app dev can invent those types
- records have URIs
- records are signed so cacheable without trust
- relay aggregates everyone’s events into a global stream
- backends subscribe to relay and update local DBs
- you can host your data
- it's just json
- it's typed but any app dev can invent those types
- records have URIs
- records are signed so cacheable without trust
- relay aggregates everyone’s events into a global stream
- backends subscribe to relay and update local DBs
November 1, 2024 at 12:02 AM
atproto aha moments:
- you can host your data
- it's just json
- it's typed but any app dev can invent those types
- records have URIs
- records are signed so cacheable without trust
- relay aggregates everyone’s events into a global stream
- backends subscribe to relay and update local DBs
- you can host your data
- it's just json
- it's typed but any app dev can invent those types
- records have URIs
- records are signed so cacheable without trust
- relay aggregates everyone’s events into a global stream
- backends subscribe to relay and update local DBs
Oh. After many long years... HTTP now officially has a standard QUERY method. Think: cacheable, idempotent GET that can carry a meaningful payload. We can now all stop bastardizing POST.
auth48-transition.rfc-editor.org/authors/rfc1...
auth48-transition.rfc-editor.org/authors/rfc1...
auth48-transition.rfc-editor.org
June 14, 2026 at 12:40 PM
Oh. After many long years... HTTP now officially has a standard QUERY method. Think: cacheable, idempotent GET that can carry a meaningful payload. We can now all stop bastardizing POST.
auth48-transition.rfc-editor.org/authors/rfc1...
auth48-transition.rfc-editor.org/authors/rfc1...
Looks like Chromium is about to change how prefetches behave, so that only cacheable resources will be served from the prefetch cache: groups.google.com/a/chromium.o...
If you're prefetching your HTMLs, make sure they are actually cacheable, even with a short lifetime.
If you're prefetching your HTMLs, make sure they are actually cacheable, even with a short lifetime.
Intent to Ship: removing the five-minute rule for <link rel=prefetch>
groups.google.com
November 28, 2024 at 10:31 AM
Looks like Chromium is about to change how prefetches behave, so that only cacheable resources will be served from the prefetch cache: groups.google.com/a/chromium.o...
If you're prefetching your HTMLs, make sure they are actually cacheable, even with a short lifetime.
If you're prefetching your HTMLs, make sure they are actually cacheable, even with a short lifetime.
- non-personalized components are globally cacheable by output hints, invaliadted by time or at uris / backlinks
- personalized components are non-cacheable, indicated by a flag
- components that take children always receive opaque nodes, which allows caching shells whose contents updates faster
- personalized components are non-cacheable, indicated by a flag
- components that take children always receive opaque nodes, which allows caching shells whose contents updates faster
January 29, 2026 at 12:11 AM
- non-personalized components are globally cacheable by output hints, invaliadted by time or at uris / backlinks
- personalized components are non-cacheable, indicated by a flag
- components that take children always receive opaque nodes, which allows caching shells whose contents updates faster
- personalized components are non-cacheable, indicated by a flag
- components that take children always receive opaque nodes, which allows caching shells whose contents updates faster
With the cacheable image so they can geolocate you 👍
January 23, 2025 at 12:21 AM
With the cacheable image so they can geolocate you 👍
🚨 An npm worm is spreading live, while half of the security industry is at #BlackHat in Vegas. talk about timing!
@socket.dev is now tracking 2,234 malicious package artifacts across 444 unique packages in the keyv/cacheable compromise.
Average detection time: 5 min 18 sec after publication
@socket.dev is now tracking 2,234 malicious package artifacts across 444 unique packages in the keyv/cacheable compromise.
Average detection time: 5 min 18 sec after publication
Popular npm Packages in the keyv and Cacheable Namespaces Co...
Popular npm packages keyv and cacheable compromised.
socket.dev
August 4, 2026 at 2:46 PM
🚨 An npm worm is spreading live, while half of the security industry is at #BlackHat in Vegas. talk about timing!
@socket.dev is now tracking 2,234 malicious package artifacts across 444 unique packages in the keyv/cacheable compromise.
Average detection time: 5 min 18 sec after publication
@socket.dev is now tracking 2,234 malicious package artifacts across 444 unique packages in the keyv/cacheable compromise.
Average detection time: 5 min 18 sec after publication
👀 New HTTP QUERY method
Similar to GET, with a body allowed
Safe, idempotent, cacheable
There are also headers to:
- give the query its own URI (Location) to query again without body
- give the response its own URI (Content-Location) to get the same response again
www.reddit.com/r/webdev/com...
Similar to GET, with a body allowed
Safe, idempotent, cacheable
There are also headers to:
- give the query its own URI (Location) to query again without body
- give the response its own URI (Content-Location) to get the same response again
www.reddit.com/r/webdev/com...
July 7, 2026 at 9:10 AM
👀 New HTTP QUERY method
Similar to GET, with a body allowed
Safe, idempotent, cacheable
There are also headers to:
- give the query its own URI (Location) to query again without body
- give the response its own URI (Content-Location) to get the same response again
www.reddit.com/r/webdev/com...
Similar to GET, with a body allowed
Safe, idempotent, cacheable
There are also headers to:
- give the query its own URI (Location) to query again without body
- give the response its own URI (Content-Location) to get the same response again
www.reddit.com/r/webdev/com...
Finally, TanStack Start now supports React Server Components!
Start's RSCs are a truly fetchable, cacheable and composable primitive that work with your favorite tools instead of dictating your entire architecture.
Oh, and one more thing... "Composite Components" 😉
tanstack.com/blog/react-s...
Start's RSCs are a truly fetchable, cacheable and composable primitive that work with your favorite tools instead of dictating your entire architecture.
Oh, and one more thing... "Composite Components" 😉
tanstack.com/blog/react-s...
React Server Components Your Way | TanStack Blog
RSCs are genuinely exciting — smaller bundles, streaming UI, moving heavy work off the client — but existing implementations force you into a one-size-fits-all pattern. What if you could fetch, cache,...
tanstack.com
April 13, 2026 at 4:37 PM
Finally, TanStack Start now supports React Server Components!
Start's RSCs are a truly fetchable, cacheable and composable primitive that work with your favorite tools instead of dictating your entire architecture.
Oh, and one more thing... "Composite Components" 😉
tanstack.com/blog/react-s...
Start's RSCs are a truly fetchable, cacheable and composable primitive that work with your favorite tools instead of dictating your entire architecture.
Oh, and one more thing... "Composite Components" 😉
tanstack.com/blog/react-s...
HTTP gets a QUERY method so complex searches can stop pretending to be POST
HTTP gets a QUERY method so complex searches can stop pretending to be POST
New verb carries request content while remaining safe, idempotent, and cacheable
www.theregister.com
July 13, 2026 at 11:15 AM
HTTP gets a QUERY method so complex searches can stop pretending to be POST
We have plans here. Server functions will be cacheable when you express this caching intent. Last Next.js conf we teased “use cache” which we are getting ready to release soon. It won’t alone deliver what you are looking for but it’s a building block that will allow the next step soon after
April 26, 2025 at 8:39 PM
We have plans here. Server functions will be cacheable when you express this caching intent. Last Next.js conf we teased “use cache” which we are getting ready to release soon. It won’t alone deliver what you are looking for but it’s a building block that will allow the next step soon after
If I were to break down the percentages of work our AppView is doing on a like, DB query throughput basis:
~60-65% of queries are for hydrating relationships between actors
~25-30% of queries are for timeline fanout
~10-15% of queries are hydrating static content (heavily cacheable)
~60-65% of queries are for hydrating relationships between actors
~25-30% of queries are for timeline fanout
~10-15% of queries are hydrating static content (heavily cacheable)
April 27, 2025 at 2:22 AM
If I were to break down the percentages of work our AppView is doing on a like, DB query throughput basis:
~60-65% of queries are for hydrating relationships between actors
~25-30% of queries are for timeline fanout
~10-15% of queries are hydrating static content (heavily cacheable)
~60-65% of queries are for hydrating relationships between actors
~25-30% of queries are for timeline fanout
~10-15% of queries are hydrating static content (heavily cacheable)
HTTP gets a QUERY method so complex searches can stop pretending to be POST
HTTP gets a QUERY method so complex searches can stop pretending to be POST
New verb carries request content while remaining safe, idempotent, and cacheable
www.theregister.com
July 14, 2026 at 11:16 AM
HTTP gets a QUERY method so complex searches can stop pretending to be POST
if the data flows down, how do you scale the network? you don’t want to query from source every time. but if you rely on a cache, how do you trust it? what if it fakes the data?
atproto solves this by having every change be signed by its author. so any slice of the network is trustlessly cacheable.
atproto solves this by having every change be signed by its author. so any slice of the network is trustlessly cacheable.
November 4, 2024 at 4:17 AM
if the data flows down, how do you scale the network? you don’t want to query from source every time. but if you rely on a cache, how do you trust it? what if it fakes the data?
atproto solves this by having every change be signed by its author. so any slice of the network is trustlessly cacheable.
atproto solves this by having every change be signed by its author. so any slice of the network is trustlessly cacheable.
Active exploits found by none other than the great and all powerful @feross.bsky.social & co
socket.dev/blog/popular...
socket.dev/blog/popular...
Popular npm Packages in the keyv and Cacheable Namespaces Co...
Popular npm packages keyv and cacheable compromised.
socket.dev
August 4, 2026 at 12:32 PM
Active exploits found by none other than the great and all powerful @feross.bsky.social & co
socket.dev/blog/popular...
socket.dev/blog/popular...
For my first post on Bluesky, I'm really happy to announce that I just released AssertJ-DB 3.0.0 with some major changes:
- Cacheable database schema metadata
- Fluent API
The cache feature should really speed up your test suite if it contains a lot of change assertions
github.com/assertj/asse...
- Cacheable database schema metadata
- Fluent API
The cache feature should really speed up your test suite if it contains a lot of change assertions
github.com/assertj/asse...
Release v3.0.0 · assertj/assertj-db
⭐ Improvements
Fluent construction of AssertJ-DB connection to database
Cacheable database schema metadata
Fluent database element builder ( Table, Request, Changes )
🚧 Breaking changes
Remove c...
github.com
November 26, 2024 at 1:35 PM
For my first post on Bluesky, I'm really happy to announce that I just released AssertJ-DB 3.0.0 with some major changes:
- Cacheable database schema metadata
- Fluent API
The cache feature should really speed up your test suite if it contains a lot of change assertions
github.com/assertj/asse...
- Cacheable database schema metadata
- Fluent API
The cache feature should really speed up your test suite if it contains a lot of change assertions
github.com/assertj/asse...
We reviewed actions/setup-java released versions v1–v5 and main, against today’s Keyv/Cacheable npm supply-chain attack.
No compromised versions or known IOCs were found in the lockfiles or shipped bundles.
This is a point-in-time assessment:
github.com/actions/setu...
No compromised versions or known IOCs were found in the lockfiles or shipped bundles.
This is a point-in-time assessment:
github.com/actions/setu...
Security review: Keyv/Cacheable npm supply-chain incident · Issue #1193 · actions/setup-java
Summary As of 2026-08-04, the current actions/setup-java dependency tree and distributed JavaScript bundles are not affected by the Keyv/Cacheable npm supply-chain attack reported by Wiz Research. ...
github.com
August 4, 2026 at 3:18 PM
We reviewed actions/setup-java released versions v1–v5 and main, against today’s Keyv/Cacheable npm supply-chain attack.
No compromised versions or known IOCs were found in the lockfiles or shipped bundles.
This is a point-in-time assessment:
github.com/actions/setu...
No compromised versions or known IOCs were found in the lockfiles or shipped bundles.
This is a point-in-time assessment:
github.com/actions/setu...
"Dynamic Client Registration itself is now formally deprecated in favor of CIMD."
blog.modelcontextprotocol.io/posts/2026-0...
blog.modelcontextprotocol.io/posts/2026-0...
The 2026-07-28 Specification
The 2026-07-28 Model Context Protocol specification is out, bringing a stateless protocol core, Multi Round-Trip Requests, header-based routing, cacheable list results, authorization hardening, a form...
blog.modelcontextprotocol.io
July 31, 2026 at 4:51 PM
"Dynamic Client Registration itself is now formally deprecated in favor of CIMD."
blog.modelcontextprotocol.io/posts/2026-0...
blog.modelcontextprotocol.io/posts/2026-0...
declarative control over prefetching github.com/vercel/next....
July 23, 2026 at 2:15 PM
declarative control over prefetching github.com/vercel/next....
(the cid cache hit rate is really around 65% (it's computing against all cids, not potentially-cacheable cids), and you can actually get it to almost 90% with a different cache approach (LRU instead of last-commit), but keeping the client's cache in sync would be way harder/messier.)
bad news and good news
intuition was right: nearly 90% of MST CIDs in the firehose are rebroadcasts of recently-emitted CIDs.
bad: an LRU cache needs to be kind of big to get a high hit ratio (128 entries = 4.5KB/repo)
good: keeping CIDs from the prev commit only yields 60–70% hit ratio
intuition was right: nearly 90% of MST CIDs in the firehose are rebroadcasts of recently-emitted CIDs.
bad: an LRU cache needs to be kind of big to get a high hit ratio (128 entries = 4.5KB/repo)
good: keeping CIDs from the prev commit only yields 60–70% hit ratio
May 1, 2026 at 6:40 PM
(the cid cache hit rate is really around 65% (it's computing against all cids, not potentially-cacheable cids), and you can actually get it to almost 90% with a different cache approach (LRU instead of last-commit), but keeping the client's cache in sync would be way harder/messier.)
no because labelers are part of the hydration/filtering post-appview. they could but it would make feeds less cacheable.
March 29, 2026 at 11:12 PM
no because labelers are part of the hydration/filtering post-appview. they could but it would make feeds less cacheable.
i think the shape of the data flow matters. appviews don't talk to each other so e.g. your post going viral on the network won't bring down your AppView (or your PDS). data from PDS's is safely cacheable and broadcastable. being able to reuse communal-run modular pieces matters
August 24, 2025 at 6:24 PM
i think the shape of the data flow matters. appviews don't talk to each other so e.g. your post going viral on the network won't bring down your AppView (or your PDS). data from PDS's is safely cacheable and broadcastable. being able to reuse communal-run modular pieces matters
Honestly wouldn't do too much in terms of infra for us. It'd help distribute points of failure which means more of the network stays alive when a single service provider has issues but the architecture is designed so hosting your own data is very cheap and incredibly cacheable on the network.
November 15, 2024 at 3:19 AM
Honestly wouldn't do too much in terms of infra for us. It'd help distribute points of failure which means more of the network stays alive when a single service provider has issues but the architecture is designed so hosting your own data is very cheap and incredibly cacheable on the network.
200s are "heuristically cacheable" but many error status codes aren't. so to avoid caching shenanigans may at least want to toss a no-store on those kinds of responses.
April 29, 2024 at 7:53 PM
200s are "heuristically cacheable" but many error status codes aren't. so to avoid caching shenanigans may at least want to toss a no-store on those kinds of responses.
Hls makes you split videos into small cacheable chunks *anyways*, its actively not-useful to have one big file for videos
November 26, 2024 at 1:54 AM
Hls makes you split videos into small cacheable chunks *anyways*, its actively not-useful to have one big file for videos