Cached version here: http://www.alohastartups.com/2014/07/12/wtf-mblooms-first-in...
Other writeups on this issue: http://www.hawaiiweblog.com/2014/07/10/mbloom-funds-mbloom http://www.civilbeat.com/2014/07/civil-bytes-is-there-a-conf...
1,391 karma · joined October 24, 2009
Cached version here: http://www.alohastartups.com/2014/07/12/wtf-mblooms-first-in...
Other writeups on this issue: http://www.hawaiiweblog.com/2014/07/10/mbloom-funds-mbloom http://www.civilbeat.com/2014/07/civil-bytes-is-there-a-conf...
That said, performance is totally on the side of lapis/openresty. The performance of my side project totally depends on how fast postgres responds to the needed queries.
I work at a pretty security conscious company (this might be an understatement, we're pretty big on security), and even as a developer on the inside I'd have to get pretty creative to get access to our production servers.
I really liked the cypher query language of neo4j, it was super easy to learn and is way simpler to express complex traversals across a graph. Writing a query in postgres to traverse a graph isn't as straightforward, and unless you're doing it often will take a long time to get used to.
>The most common case is when a free library's features are readily available for proprietary software through other alternative libraries. In that case, the library cannot give free software any particular advantage, so it is better to use the Lesser GPL for that library.
>This is why we used the Lesser GPL for the GNU C library. After all, there are plenty of other C libraries; using the GPL for ours would have driven proprietary software developers to use another—no problem for them, only for us.
>However, when a library provides a significant unique capability, like GNU Readline, that's a horse of a different color. The Readline library implements input editing and history for interactive programs, and that's a facility not generally available elsewhere. Releasing it under the GPL and limiting its use to free programs gives our community a real boost. At least one application program is free software today specifically because that was necessary for using Readline.
This is doable since dictionaries and blocks show up in different contexts. A block never follows =,:, ;, or (, so with a context sensitive grammar this can be accomplished. I guess it depends how they intend to support anonymous functions (def{}?)
So even if you can audit and fix the js code you're running, you're still including content served straight from someone else's machine.
When we did this similar thing for sopa blackout (https://github.com/sirpengi/sopablackout), our widget was entirely self-contained (and under 200 LOC). And if you didn't trust our server you could host it entirely yourself.
Tampering with the RNG probably provides the best value for an attacker, and is harder to detect.
The link that specifically shows people complaining that nobody viewed their app or looked at their video is this: http://salesforce1million.challengepost.com/forum_topics/264...
So really, the math works out such that if you fail together, you'll have X amount of downtime. If you fail seperately, you'll have X*3 amount of degraded service.
I'm not concerned about users uploading files to someone else's server, but sometimes servers are configured to rewrite certain urls, or to do redirects, and attackers might be able to take advantage of that to rack up some user's bill. For instance, some cache-busting schemes rewrite /d/_somehash_/file or /d/file._somehash_ to the same /d/file and your service has no idea they're actually the same file. In this case, some attacker can just make up random hashes to make false transcode jobs, even though they all return the same video file.
Does your service treat /d/file and /d/file?v=1 as separate files? If so that can also be used as an attack vector as someone can add whatever parameters they want to the url, and most webservers silently ignore those. You can close this by ignoring those parameters, but then you exclude sites that serve files through some script (like /cgi/get-file?f=somefile)
Does your service whitelist file extensions? Or stream in file headers first to decide whether any file is a valid movie, otherwise stopping the download? Otherwise, any site that hosts other large files (say, linux ISOs) can force you to download really large files and waste bandwidth and resources.
With that in mind, I think if you also make it so your clients must also provide a subdirectory (for example, I could set it so only files within /static/videos/ are valid, as that's where I've decided to put my videos) then most of these issues will be taken care of. Your clients would then just need to make sure they don't have rewrite enabled for that subsection of their site, and make sure other large assets live outside that subdirectory.
The problem is that this has to be done server side because the key needs to be secret (thus, this can't happen in javascript in the client), which breaks the whole idea you're selling (API-less transcoding).
You could also provide your clients with urls by having them submit the request on your site (obviously, behind a login). The user would then use that url in the src of their video tag. Again, this method goes against your selling point.
In case resolution choice is added as an option, what's to keep any random Joe from requesting videos with various resolutions (or very huge resolutions) and racking up the bill for me?
As for bulk export, yes in the sense of "download all the calculated recommendations" (perhaps you want to dump that to a local db to access that data directly, rather than rely on hitting my api every time you need something), but not "download all the data exactly as I put it in". For the most part, the data you put into my service should already exist in your own database (what products your users have purchased, what products your users elected as liking). In addition, you'll likely want to use a one-way hash to hash sources (especially if you use usernames or email addresses to uniquely identify users).
I probably won't launch with the ability to add weights to edges, although that's probably one of the first post MVP features on my todo list.