@TechieNotNetie@social.vivaldi.net

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@dansup@mastodon.social @HalleAndert@social.vivaldi.net @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed FYI, we just released a minor update to 8.0 with a patch for the Pixelfed login problem (plus a number of Chromium patches from upstream) https://vivaldi.com/blog/android/minor-update6-android-8-0/

Vivaldi Browser

Minor update(6) for Vivaldi Android Browser 8.0 | Vivaldi Browser

Today’s update includes a Chromium bump.

6

1

2

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@dansup@mastodon.social @HalleAndert@social.vivaldi.net @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed FYI, the Android team is currently investigating this issue.

2

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 5mo ago

Replying to

@pvdrijst @Linux @atoponce @Vivaldi That a feature is mentioned in that list does not mean it is enabled, it just means the code exists (and it is too difficult to actually disable this at compile-time, that requires too many patches). In fact, the feature the whole download depends on (OptimizationGuide) is disabled in Vivaldi and have been for a while, and I recently disabled the PromptApi features to be on the safe side. In addition, for a while what I suspect is the download URL for the model files being discussed has been pointed at a server that only return 404 responses . in case it should somehow be loaded.

3

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 5mo ago

Replying to

@Vivaldi@social.vivaldi.net @cloudflare@noc.social Update (Apr 15, 1900 CET): Cloudflare just pushed a fix to production, and my testing on a site seems to confirm that the problem is gone.

3

0

2

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@dansup@mastodon.social @HalleAndert@social.vivaldi.net @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed As I don't work with Android, I am not able to say; I will need to bring in somebody from the Android team for this, and most of them are on US time currently.

1

2

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net Unfortunately, there is no test case, so cannot check what the problem is.

1

17

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 8mo ago

Hello all, we are aware that Palo Alto Network's Cortex XDR, an "AI"-driven security (aka "anti-virus") application, is hallucinating and is incorrectly claiming @Vivaldi@social.vivaldi.net is dangerous. This is a false positive, and we are waiting for them to fix their product.

In the meantime, we are aware that their software apparently provides options for you to exclude Vivaldi from scanning, if your organisation allows it.

#Vivaldi #AntiVirus #PaloAltoNetworks

social.vivaldi.net

Vivaldi Social

3

0

4

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 6mo ago

Replying to

@mkljczk @Vivaldi The confusion here is that there is the good kind of statistical analysis; like weather forecasts, translations; and the contentious kind; LLMs and image generators whose content is based on ignoring copyright and wasting vast amounts of energy and the unknown privacy or security risks of agentic control. Depending on who you talk to, and depending on how much they want to jump on a buzzword wagon, these can all be called "AI", but in reality, there is a big difference between the benefits of one, and the harm created by the other. We do not allow "AI" agentic features. We do not build in LLM capabilities. Regarding the Vivaldi translation service, we host it on our own servers. Lingvanex provides the models our servers use to perform the translation, and they do not get data from our servers for training. Re the translation API, we have confirmed that parts of it works, using an installed translation component (in vivaldi:components), containing the translation model locally; but parts of the related APIs are disabled hard (which is why the playground does not work). You may also wish to note that neither translation service use a high energy consumption service. Translation is fast statistical analysis. It does not function in the same way as an LLM.

2

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 5mo ago

Replying to

@Taffer@mastodon.gamedev.place @Vivaldi@social.vivaldi.net @cloudflare@noc.social AFAICT it is supposed to be verifying that there is a human using the browser; part of it is automatic, but what it looks like is that if it hits at least some variants of "Vivaldi" in the Client Hints, it falls back to a manual "confirm you are a human" checkbox and that subsequently fails, blocking use of the site. As mentioned in the post, we are trying to get in touch with Cloudflare about this problem.

1

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 5mo ago

Replying to

@missingno@social.vivaldi.net @Vivaldi@social.vivaldi.net The history of this sequence of lies about which browser is being used is covered in an article linked from the "Client Lies" article linked from my forum post. As a shortcut it is https://humanwhocodes.com/blog/2010/01/12/history-of-the-user-agent-string/ TL;DR: MSIE started it

Human Who Codes

History of the user-agent string - Human Who Codes

The Official Web Site of Nicholas C. Zakas

1

3

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 6mo ago

Replying to

@mkljczk @Vivaldi The feature is not included in Vivaldi. That references to various AI related features can be found around the browser does not mean they actually _work_, and if you check the translation playground in the page you linked to, you will find that it does not work (compare to the other browser). In this case, AFAICT from the related pages to the page you linked to, that feature requires that a component called "Optimization Guide" in Chromium works, and that is one of the AI-related components where we have disabled important parts of it (we concentrate on disabling the choke points), including that the download URL has been changed to point at one of our servers that only return "404 Not Found", thus breaking the model downloads.

1

4

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed

There may be a third possible cause of the problem: Vivaldi disables the Chromium builtin Gemini LLM (so-called AI) system hard at build time (other related systems are run-time disabled); I don't know what Chromite does, but if it also disabled Gemini, then yet another possibility is that something in the authentication flow tries to use Gemini-related APIs and that Vivaldi's failure to "support" it breaks the process.

0

4

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed Anyway, I did a bit of code search, this seems to involve the Auth Tab (part of Custom Tabs) in Android Chromium

0

9

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net In any case, I'd like more information about the URL(s) and their format. That is where the "real action" is. Also: Did this work before? If so, when did it break?

0

14

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net I think I need a bit more detailed information about how you proceed along that sequence. E.g what is "the app"?

0

15

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed It is of interest to know if it worked earlier in v8.0, of if "worked in the past" means 7.9 or earlier. The question is if this worked recently, that is, within the last few months?

0

12

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed OK, that makes it difficult to find out what broke; it is also a authentication flow I am not familiar with. However, what I would have started with is to make sure what was sent from the server to Vivaldi included all parameters in correct format (e.g. if everything should be in the URL, then some of the data should probably not be sent as payload content). Then I would have looked at what arrived at the app.

0

10

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed Hmm, Chromium 143 is a bit on the old side. Which Chromium version is Chromite based on?

0

7

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 3mo ago

Replying to

@HalleAndert@social.vivaldi.net @dansup@mastodon.social @ruario@social.vivaldi.net @jon@social.vivaldi.net @pixelfed

Thanks for that info. So current version of both Vivaldi and Chromite are based on Chromium 148, while the Chrome installation referenced above is no doubt based on Chromium 149.

That suggest two possibilities:

There was a bug related to this introduced in Chromium sometime between v146 branch-off and v148 branch-off, that was fixed in v149.

or

The authentication used by Pixelfed performs browser sniffing and "refuses" to provide the complete and correct redirect query URL to pre-v149 browsers.

Considering that possibility 1 would have meant that Chrome 147 and/or Chrome 148 would have broken the same way as Vivaldi and Chromite, I cannot imagine that the Chromium team would not have backported an urgent fix to the affected release(s) in a minor update, which would at least have fixed Vivaldi.

So, my guess is that this is server- or JS-script-side browser sniffing, which means this is something that Pixelfed will have to fix.

0

5

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 5mo ago

Replying to

@missingno@social.vivaldi.net @Vivaldi@social.vivaldi.net And when I and others in Vivaldi worked at Opera, we had serious problems with web site compatibility because of it.

0

1

0

TechieNotNetie @TechieNotNetie@social.vivaldi.net

· 4mo ago

Replying to

@missingno@social.vivaldi.net @hrx@mastodon.online The Client Hints Branding setting DOES NOT change the User Agent header, only the Client Hints Brands header.

0

1

0