vrtx0

Absolutely. Truly innovative research and projects are always fascinating to me. But even generalized to AI/ML, I can’t think of anything more recent than GANs (2015), Transformers (2017), and maybe AlphaFold 2 (2021)…

Just my personal taste though; I don’t mean to knock LLMs.

So, I’m real, and just trying to offer constructive feedback for a few errors I believe I noticed.

I could be wrong though —- could you be specific? I don’t want to misinform anyone…

I provided a copy/paste from the site about the envp array size you asked about.

I clarified why I mentioned fork().

I tried to explain the difference between registers and variables.

I’m not trying to show off or bring anyone down… I just like to help people. I’m old (my first Linux kernel commit was in 2004). And I could be wrong — please LMK if I made a factual error (I’d appreciate it, honestly).

All good?

x20 is a general purpose register; optimizing compilers can use it for any number of variables, immediate values or intermediate computations at different points within that same function — or none at all (the variable ep could be optimized away).

Re: fork(), I just meant to be thorough in explaining the environment is copied, not shared by processes. Setenv() only affects the process from which it’s called.

The array size bit in the article: The value 0x220 looks suspiciously close to the size of the old environment in 64-bit words (0x220 / 8 = 68), and this value was written over the terminating NULL of the environment block…

HTH!

Let me try to help:

1. If a process crashes and dumps, be sure to look at the system log of the cause (e.g. SIGSEGV, OOM, invalid instruction, etc.)

2. Be certain you’re looking at the right core dumps — I believe UID 1000 just means posix UserID (which is unrelated to a PID), though I don’t use containers.

3. Stay focused on the right level of abstraction — memory model details are great to know, but irrelevant here.

4. Variables do not correlate 1:1 with registers, except in C calling conventions. The assumption about x20 and a local variable is incorrect, unfortunately.

5. getenv() and setenv() do not work as implied in the post. When a process starts via execve(), the OS/libc constructs a new snapshot of the environment, and cannot be modified by an ancestral process. It’s a snapshot in time, unless updated by the process itself. When a process fork()s, the child gets a new copy of the parent’s environment — updates do not propagate.

getenv() is thread safe and reentrant. You don’t use an environment to pass shared data — setenv() is generally used when constructing the environment for a child process before a fork(). See man environment.

6. FWIW, ‘char** env’ is a null-terminated array of pointers, so dumping memory from *env (or env[0]) is only valid until you hit the first NULL. The size of the array is not stored in the array.

I hope this helps! And apologies if this is redundant — I read so many comments; mostly variations of “the problem with getenv is x”, but gave up before reading all of the (currently) 168 comments.

I don’t see any merit, honestly. That would assume one is able to audit every bit of code they run, including updates, and control the build system.

I mean, the Wally paper contains enough information to effectively implement homomorphic encryption for similar purposes. The field was almost entirely academic ~12 years ago…

I miss talking shop on HN. Comments like that are why we can’t have nice things.

Sorry, what do you mean by “proprietary details”?

Do you mean the ability to search in Apple Photos is “privacy-bruising”, or are you referring to landmark identification?

If the latter, please note that this feature doesn’t actually send a query to a server for a specific landmark — your device does the actual identification work. It’s a rather clever feature in that sense…

Thank you! This is exactly the information the OP seems to have missed. It seems to confirm my suspicion that the author’s concerns about server-side privacy are unfounded — I think:

> The client decrypts the reply to its PNNS query, which may contain multiple candidate landmarks. A specialized, lightweight on-device reranking model then predicts the best candidate…

[please correct me if I missed anything — this used to be my field, but I’ve been disabled for 10 years now, so grain of salt]

Is the Apple Photos feature mentioned actually implemented using Wally, or is that just speculation?

From a cursory glance, the computation of centroids done on the client device seems to obviate the need for sending embedded vectors of potentially sensitive photo details — is that incorrect?

I’d be curious to read a report of how on-device-only search (using latest hardware and software) is impacted by disabling the feature and/or network access…

Whoa, very biased article (especially for LWN). Only cites media coverage; no links supporting that Amazon, MSFT, Google, etc. were in fact EEE’ing (or at best, behaving unethically) with each of these projects.

It even suggests cloud providers did contribute, and uses bad data (git commits “by employer” w/o dataset) that basically contradicts their argument.

I may be biased, as I saw Amazon doing exactly what this article claims “maybe they weren’t”. But statements like this seem intentionally misleading, and easily disproven:

“Distributing a source-available version of MongoDB could be seen as a loss-leader strategy to reach developers that the company wagered did not care about open-source.”

MongoDB is still “source-available”, and on the same GitHub repo I’ve used since 2010. The SSPL only impacts cloud-providers, and has exceptions for cloud providers who release their source code.

The OSI doesn’t get to define open-source. Neither do I, but at least I was part of the community for ~20 years…

I think “including, without limitation, […]” applies to the breadth of components, not depth, right? I mean, I’m not a lawyer, but that seems to be what syntactic context and logic indicate… no?

If you disagree, could you indicate the relevant text?

Er, I could be wrong, but I think you’re missing the scope set in the first sentence:

If you make the functionality of the Program or a modified version available to third parties as a service, you must make the… SNIP …programs that you use to make the Program or modified version available as a service…

I’m eliding for clarity, but the NAS doesn’t make the program available as a service. The code that accesses the file system on the NAS to offer your service? Probably need to release code that calls fread/fwrite/NtFileX in your infrastructure code.

I get that it sounds vague and everything, but the FAQ also clarifies none of this applies unless you’re competing and targeting third parties. If you’re one of the few companies who want to do that, your legal team can formalize the line of demarcation.

Apologies, I hate defending the SSPL, but I can’t think of any better way to stop the monopolistic and EEE practices against open source projects. If anyone has a better solution to protect the freedoms of open-source developers, please, please publish it!

The cloud provider wouldn’t have to, if you are the one running your infra. The SSPL restrictions only apply to businesses that offer MongoDB as a service.

In fact, you can build a similar service and offer it within your organization (and subsidiaries), and you still don’t have to release anything. The license only applies to companies like Amazon if they offer MongoDB as their own service (DocumentDB).

I know that’s a bit tangential, but hope that helps a bit?

No, full list is in the license, but it’s only those components that they provide MongoDB as a service to end users. Doesn’t penetrate OS abstractions AFAICT, but I’d check the license and/or consult an expert if starting a relevant business!

FWIW, just realized it doesn’t even apply if deployed internally (within an organization and/or subsidiaries)…