soamv

Yes, this is very much on the roadmap. We're working on better support for compiled languages. Once we have that, adding a language should be a matter of creating a plugin.

(You watch this issue if you'd like to stay up to date on progress https://github.com/fission/fission/issues/24)

[fission developer here]

About that 100msec overhead -- there's a bunch of optimizations yet to be done, and also we think we can reduce the median at the expense of tail latency; but for now 100msec is good enough performance for a good enough set of use cases (but of course we'd love to hear from people for whom it really doesn't work at all).

Also, we could add knobs that let you tune how long a function stays "warm", so you can tune the latency overhead vs. resource consumption trade-off.

> But with modern progress in fast-starting compiled languages like Go and Rust, I'd love to have some easy way to supply small static binaries instead of source code.

We're very interested in this use case. We're redesigning the function environment stuff right now to better support compiled languages, and allowing users to bypass the source-code level stuff is quite possible.

BTW, to be clear, for a compiled language Fission wouldn't compile source code at runtime; there would be separate build stage that would run when you add the function to fission. I'm also not sure why you say we pay interpreter startup overhead on every request (we don't).

Hi all, Fission developer here. Happy to answer any questions.

The comment you're replying to doesn't say that etcd is insecure; just that access to etcd is equivalent in power to Kubernetes API access.

It's hard to have a productive discussion in such a confrontational tone, especially when you don't really have all the facts. If you have questions about how to secure Kubernetes communication with etcd, feel free to ask!

Nice work. Feature request: track CFP dates and let people find conferences with upcoming CFP deadlines.

BTW, you have "USA" and "United States" as two countries, might want to combine those.

Yeah. This is an important problem with Lambda that doesn't get nearly enough attention.

Ok, but is it actually computing the same thing? Are the blurred images from photopea and photoshop identical? The article mentions always doing no more than N*10 steps -- is this an approximation algorithm?

From my experience hosting a bunch of user-provided images:

1. Strip all metadata but provide downloads of originals somewhere

2. Keep it simple, just use imagemagick's convert to remove profiles (but don't use imagemagick for file type detection)

3. If the image has orientation exif tags, rotate the image to the right orientation (-auto-orient) before removing the exif profile.

4. Don't remove image profile data. Or convert to sRGB first.

There are financial benefits to being able to very easily overcommit resources (although Lambda is just one way to do that).

I run https://glimmer.cc. It pays for its own hosting costs. It's not very profitable but it's steadily growing.

Nice! Just tried it out and it seems to work well.

Why does dumb-jump-go want me to save my files before jumping?

I read the article and the linked blog and I still can't tell what this has over plain Kubernetes deployed on AWS -- what am I missing?

> Storage for applications deployed on Supergiant is allocated on high-availability arrays, however, it does not use Kubernetes' persistent volumes. Currently available on AWS, Supergiant uses Amazon's Elastic Block Store (EBS) so that volumes can move freely around the cluster.

Kubernetes has EBS support, right? What does Supergiant do better here?

I haven't yet figured out exactly how much to offer, but something like 10GB is what I'm thinking.

This is a great question! I find that the act of explaining the problem itself helps a great deal in figuring out what the pieces are.

It's totally okay to just talk to a developer about the problem and start sketching out your ideas, without necessarily knowing the specific tasks that you're going to assign.

The abstraction level of the tasks that you should assign will also come out of the level of discussion you're able to have with the person. Everyone has different abilities/amounts of domain knowledge/available time/etc., so there is never "One True Task List" for implementing something; it must depend on the developer.

Finally, it is deeply motivating for the person who's going to do the work to be an involved part of the planning discussion. It's important to anyone to have a certain amount of autonomy in their work, and making them be a part of planning is a great way to do that.

Well, this is about userspace access to nvram, with the nvram mapped as memory. It just so happens that cache management is one of the hard parts of doing that, so that's what these new instructions are for.

You started with asking what you can do to help, then went on to discount her experience as an outlier despite her clearly saying it wasn't, and then finished more or less by implying that _she's_ racist.

So if you're asking what you should do, the answer is: the exact opposite of whatever you're doing now.

Or, A cuts the pizza down the middle, B chooses a half, A and B cut their halves into thirds and C chooses one third each from A's and B's halves.

That's interesting. What is the cost of switching the stack when calling C code? Do you allocate a new stack every time a C function is called from Rust?

If you inline a compiled C function, don't you have to do the stack switching for whatever that inlined function might call? Or do you only inline code that doesn't have any function calls?

Part 1: http://irinawerning.com/back-to-the-fut/back-to-the-future/

Can't you just do:

  p/x *foo@100

Compile time asserts by expanding a macro to a switch-case statement.

Actually, amazon hasn't been very clear about whether or not physically lending a kindle is within their terms of service.

http://www.libraryjournal.com/article/CA6649814.html?rssid=1...

Can you actually tell us a bit about the new DB you're building? (What is it designed to run on, what kind of failures are you aiming to tolerate, and so on.)

Check out ruboto-core at http://ruboto.org/

I see that lispy is using instance_exec -- which means that the scope of code inside that code is the lispy instance, which might not be what the programmer expects. For example if he does this inside a method he'd find his instance variables missing. (Is that right, or am I misreading the code?)

You could use some sort of global or thread local variable to maintain the context, and then you wouldn't lose the instance scope, but then this code will not mix well with yield (and continuations and fibers etc.)

So then it looks like the only way to solve this problem really neatly is in a language that has macros.

(I made some attempts at this problem too, at https://github.com/soam/blox . It is slightly more complex than lispy; see for example https://github.com/soam/blox/tree/master/samples/webservice .)

Knuth's Mathematical Typography is interesting too: http://projecteuclid.org/DPubS?service=UI&version=1.0...

Is the average number of comment upvotes per day the same? Would that prove that people don't care about how many upvotes a comment has when they vote?

Or how about the distribution of upvotes over comments under one post -- are upvotes more uniformly distributed?

Do the comments on the bestcomments list have greater or fewer upvotes on an average?

I have also not yet read the paper, but to your point about SSDs -- SSDs are definitely good for random reads, but for random writes, it seems like having an append-only structure would help avoid write amplification. If you do a lot of small random writes to an SSD that will be slow and reduce the disk's life drastically.

(That said, I'm no SSD expert, so I might be way off)

Hmm, that looks like a different paper. Try this: http://arxiv.org/PS_cache/arxiv/pdf/1103/1103.4282v2.pdf

It seems that both reddit and metafilter seem have stronger meta discussions than HN. Reddit seems to have meta posts on the frontpage every once in a while, while metafilter has a fulltime forum (metatalk) dedicated to meta discussions.

Though there are meta discussions once in a while on HN too, they tend to be more general in nature, not specific to a certain comment or post.

I think an active meta discussion community would help with continuous small corrections, and eventually improve people's opinions on what kind of comments are good or bad.