rafael's Avatar

rafael

@rafael.my

software developer @bsky

510
Followers
42
Following
153
Posts
10.09.2024
Joined
Posts Following

Latest posts by rafael @rafael.my

Some specifics: I can find blob bafkreiauwb34hdu3zjvuwqrddnvee3deousrop4q257s2eqlsurpsj5uoy, which is of an image that is showing correctly, but can't from bafkreibysh2de4wg6rbcwzajgh3hwgorljluph5jfqbu6cl6p64n6vkzba. But listing blobs, both CIDs are returned.

06.03.2026 23:32 ๐Ÿ‘ 2 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0

I investigated a bit, it seems your PDS is failing to return some of your images. It will list all the blobs, but not be able to return all of them.

I'd recommend either waiting a bit if you migrated recently, or contacting the person who hosts your PDS. I can give specifics.

06.03.2026 23:31 ๐Ÿ‘ 2 ๐Ÿ” 0 ๐Ÿ’ฌ 4 ๐Ÿ“Œ 0

How long ago did you migrate to Eurosky? There might be some issue there, or still an ongoing migration issue.

06.03.2026 23:22 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 2 ๐Ÿ“Œ 0

Some images by @bergcathyb.eurosky.social load, some give a 404. It seems just a coincidence that the banner and avatar fail.

Adding `@jpeg` doesn't make them work for me.

06.03.2026 23:11 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 2 ๐Ÿ“Œ 0

Will take a look, sorry about that. This is the first issue I'm hearing about.

My best guess rn is that it is a caching issue, and it will fix itself in 24 hours.

Note that we're not deleting anything, so the original image is still there, untouched. Another guess is that it fails to convert.

06.03.2026 22:55 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

Each one copes differently

06.03.2026 15:09 ๐Ÿ‘ 5 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0

I won't deal with any of this shit on my OOO day, you can be sure of that.

06.03.2026 15:00 ๐Ÿ‘ 11 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

You had me in the first half there, ngl

06.03.2026 14:43 ๐Ÿ‘ 4 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

And also removing the format at the end of the URLs. So the current `@jpeg` will be completely gone, instead of doing `@webp`.

06.03.2026 14:27 ๐Ÿ‘ 5 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0

We're rolling out a change in image format.

If you see some images taking a second to load today, that's expected.

06.03.2026 13:47 ๐Ÿ‘ 51 ๐Ÿ” 10 ๐Ÿ’ฌ 4 ๐Ÿ“Œ 1

You, we can have alts for those things

05.03.2026 18:24 ๐Ÿ‘ 2 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

Hey devs! Heads up, we're planning on making some changes to our image CDN tomorrow.

- Images will be served as WebP by default
- The URL format you get from the Bluesky API will change a little

You can look forward to an announcement about image quality soon ๐Ÿ‘€

04.03.2026 18:32 ๐Ÿ‘ 297 ๐Ÿ” 43 ๐Ÿ’ฌ 29 ๐Ÿ“Œ 32

I think it is back now

04.03.2026 11:59 ๐Ÿ‘ 2 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

Should be back, catching up on some queues, so maybe it takes a minute, but should be good from now on (someone reading this might assume I know what I'm talking about, which is not accurate).

26.02.2026 20:00 ๐Ÿ‘ 5 ๐Ÿ” 1 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

We're investigating, sorry about that

26.02.2026 19:39 ๐Ÿ‘ 5 ๐Ÿ” 0 ๐Ÿ’ฌ 2 ๐Ÿ“Œ 4

Enjoy!

25.02.2026 00:19 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

The only other place I think it exists is in getting your own likes. Go to your profile, then Likes tab

14.02.2026 02:26 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

We call it c plus proto

13.02.2026 23:02 ๐Ÿ‘ 2 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0

Good news: @calabro.io will possibly roll out a fix for this today or next week.

13.02.2026 21:49 ๐Ÿ‘ 5 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

That's gold for when I'll work on this in the next few weeks, thanks!

13.02.2026 18:04 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

Then that's not quite what I expected. Do you know at which iteration that happened? Maybe that happened at ~290?

13.02.2026 17:31 ๐Ÿ‘ 0 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

Don't get me wrong, I'm not saying that's a nice behavior. I'm just saying that's how my mental model of the bug works right now.

13.02.2026 17:24 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

Did you confirm whether the loop is really infinite? My guess is that the cursors keep changing, and it stops after about ~290 calls, when the cursor comes back empty.

13.02.2026 17:23 ๐Ÿ‘ 0 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

> but it is not normal that many folks are unaware
Should have been
> but it is normal that many folks are unaware

13.02.2026 16:44 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0

implementation, but we actually serve production traffic from the Go dataplane. And that's where the bug lives.

Good news is that we've reached a consensus internally and we have a path forward to fix that bug (which similarly exists elsewhere), I expect to work on that within a month or so. (3/3)

13.02.2026 16:40 ๐Ÿ‘ 3 ๐Ÿ” 0 ๐Ÿ’ฌ 3 ๐Ÿ“Œ 0

implement a more performant dataplane. This is a closed-source, for many reasons (although there is a long-term intention of making it open-source once we address many concerns). This is a Go system, backed by a different database. So we support the TS dataplane as a viable reference (2/3)

13.02.2026 16:40 ๐Ÿ‘ 4 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

You might lack one piece of context. We have 2 dataplanes. This was discussed publicly in the past, but it is not normal that many folks are unaware. We started with the open-source TS dataplane that you are referring to. Over time (this precedes me), we hit scalability limitations and had to (1/3)

13.02.2026 16:40 ๐Ÿ‘ 2 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

Hey Kuba, we're aware of that. Unfortunately, that leaks some internal implementation details we don't really like and will have to change at some point.

13.02.2026 14:41 ๐Ÿ‘ 3 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0
Demonstration of where the Germ DM button appears on your Bluesky profile after you pair your accounts within the Germ DM app.

Demonstration of where the Germ DM button appears on your Bluesky profile after you pair your accounts within the Germ DM app.

we just rolled out an experimental integration with @germnetwork.com, an E2E encrypted chat app that integrates with atproto.

Germ users can add a button to their profile so others can message them from Bluesky.

we're exploring ways to connect your profile with other apps in the Atmosphere.

12.02.2026 19:23 ๐Ÿ‘ 738 ๐Ÿ” 197 ๐Ÿ’ฌ 35 ๐Ÿ“Œ 62
Preview
Permissioned Data Diary 1: To Encrypt or Not to Encrypt - Daniel's Leaflets The first in a series of posts about major design decisions along the way to a permissioned data protocol for atproto.

This explains it very well dholms.leaflet.pub/3meluqcwky22a

12.02.2026 14:45 ๐Ÿ‘ 0 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0