![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
Note to self: time to start figuring out my privacy policy vis a vis the government.
For the most part, privacy is going to be pretty clearly delineated in CommYou, same as in LJ: you say who can read your conversations, and I do my best to enforce those restrictions. But in the post-Patriot Act world, CommYou's relationship with the US government is potentially complicated, especially if I develop any real amount of international membership. Or, to put it more concretely: I need to figure out how I respond if the FBI shows up and demands access to somebody's private conversations.
With any luck, things will begin to get a little more rational after the election -- in particular, I have some hope that a Democratic president might at least put real teeth back into FISA oversight, so I can demand a clear court order and make that demand stick. But I probably have to assume that things have fundamentally changed enough that, even with the Republicans out of office, some of the climate of fear (and consequent governmental invasions of privacy) will continue.
Gah: really not stuff I want to be thinking about. And I probably have at least six months after launch before CommYou matters enough for the government to even notice its existence. But I need to be pondering this now, before it becomes A Problem, because the panic after getting a call from the FBI is *not* the time to be designing well-thought-out policy...
For the most part, privacy is going to be pretty clearly delineated in CommYou, same as in LJ: you say who can read your conversations, and I do my best to enforce those restrictions. But in the post-Patriot Act world, CommYou's relationship with the US government is potentially complicated, especially if I develop any real amount of international membership. Or, to put it more concretely: I need to figure out how I respond if the FBI shows up and demands access to somebody's private conversations.
With any luck, things will begin to get a little more rational after the election -- in particular, I have some hope that a Democratic president might at least put real teeth back into FISA oversight, so I can demand a clear court order and make that demand stick. But I probably have to assume that things have fundamentally changed enough that, even with the Republicans out of office, some of the climate of fear (and consequent governmental invasions of privacy) will continue.
Gah: really not stuff I want to be thinking about. And I probably have at least six months after launch before CommYou matters enough for the government to even notice its existence. But I need to be pondering this now, before it becomes A Problem, because the panic after getting a call from the FBI is *not* the time to be designing well-thought-out policy...
(no subject)
Date: 2008-03-27 02:12 pm (UTC)(no subject)
Date: 2008-03-27 02:18 pm (UTC)(no subject)
Date: 2008-03-27 02:23 pm (UTC)In the world we wish to create, in the world we wish we lived in, everyone would enjoy a right to privacy.
This service exists in and is governed by the laws of this world: the laws of nation-states, the laws of science, the laws of social interaction, the laws of common-sense, the laws that we like and the laws that we hate.
Both you and this service are at the heartless mercy of those laws and those who write and interpret them.
If you post something here, someone will be able to read it. We cannot guarantee that that person will have your best interests at heart. They may work for our government, for another government, for a tabloid, or it may be some bored kid with a computer that is worth more than he has earned up to this point in his short life. It might even be a "friend" who isn't, really.
If you threaten or attempt to hurt someone, you waive all pretense to a right to privacy in our eyes. If you trade in or conduct business that some jurisdiction that is touched by the Internet considers unlawful or within their power to regulate, it is possible, even probable, that they will attempt to compel us to cooperate with them. This may be done according to what you or I consider to be our inalienable rights--or not.
We did not create this service with the intent of depriving you of privacy. Our intentions are totally opposite: we wish to give you the level of privacy you wish to enjoy.
But we're not going to lie to you, so please don't come crying to us when we are pushed aside and rendered irrelevant in matters of privacy. We can't help you: we're in the same boat.
Enjoy your stay with us!
(no subject)
Date: 2008-03-27 02:50 pm (UTC)(no subject)
Date: 2008-03-27 02:54 pm (UTC)I considered beginning with ,"I thought we would start this Privacy Policy 'by taking some time out of our daily lives to sit down and have a little chat.'"
But that would have been a bit much. :P
(no subject)
Date: 2008-03-27 02:55 pm (UTC)(no subject)
Date: 2008-03-28 02:03 pm (UTC)(no subject)
Date: 2008-03-28 02:25 pm (UTC)Bit my tongue, though. Ow.
(no subject)
Date: 2008-03-27 02:24 pm (UTC)Too bad you inherently can't take the librarian tack, and not ever have any information on hand to surrender...
(no subject)
Date: 2008-03-28 02:10 pm (UTC)Well, keep in mind that accountability isn't really the issue here. I suspect that CommYou can make a reasonable case for common carriage if I'm careful about it, so I'm probably won't be held *responsible* for what users say. (Although that's still a fraught and uncertain subject.)
The real issue is government snooping. They have been installing backdoors in communication systems where they can -- the legality of that is a bit dubious, but I suspect that they've been exerting a lot of pressure when possible.
More importantly, there's a fairly big question of how free the FBI is to simply demand that communication records be handed over to them, and arrest anyone who doesn't comply. The administration has certainly been taking a hard line on this; again, the legality is sometimes tenuous, but they have enough justification in the Patriot Act to make a lot stick. I *despise* that on principle -- I'm willing to hand over records when faced with a court order, but anything less than that bugs the hell out of me...
(no subject)
Date: 2008-03-27 02:24 pm (UTC)stealseize your computers and retroactively tell you about it.My thoughts on this include making the chain of weak links visible. "All administrators to this system can see your posts, but promise by contract not to look. The hosting site can see your stuff; see their policy [here]. The government can order any of us to disclose; details [here]. Finally, assuming the site might one day be compromised, [here]'s what we do about it (encryption, backups, etc.)."
Unfortunately I do not anticipate the Patriot and attending acts to disappear any time soon, not unless we re-elect most of Congress. It's not like the act passed by a small, partisan margin.
(no subject)
Date: 2008-03-28 02:14 pm (UTC)True, although I think there's a fair amount of buyer's remorse there -- a lot was done in the heat of the moment that people have regretted since. So there is some hope of gradually weakening the more onerous and stupid bits of the act, although I don't hold out much hope of returning to the status quo ante any time soon...
(no subject)
Date: 2008-03-28 02:24 pm (UTC)Heh. Interesting way to phrase it. :)
(no subject)
Date: 2008-03-27 02:38 pm (UTC)(no subject)
Date: 2008-03-28 02:26 pm (UTC)True. Indeed, this relates to one of the stories I'm pondering: whether to provide at least an encrypted wire channel for end users. I don't want to encourage that in most cases -- for most people, for most purposes, it's expensive overkill -- but there are some legitimate and common use cases I can see for it. (In particular, when using the service through suspect wifi channels such as at a Starbucks.) We'll see on that one.
But on the main point: yes, you're undoubtedly correct. But it does stick in my craw, and I'd like to understand the degree of "no choice" I actually have. The telecomms snooping affair illustrates what I'm thinking about -- some services clearly played along, some apparently didn't. This suggests to me that the government is using pressure and influence as well as clear legal authority, and I'd like to have some idea where the line is between those, so that I can't simply be used.
Fortunately, this is a long-term problem unless I am *far* more successful than I realistically expect. But it's something worth pondering...
(no subject)
Date: 2008-03-28 02:52 pm (UTC)It is also worth considering that, if you provide such encryption tools or the embedding of encrypted materials, you'll never know if bad guys are using it too. (Define bad guy for yourself - bad is subjective.)
I suppose you could encrypt logs and other persistent data, and that you could make a good argument for doing so.
If you were building something like the old eRoom tool, or other workplace tools, then I could see how end-to-end confidentiality would be critical. Do you have such a model as your target? Or is this more "Whatya doin'", "Nuthin", "Wanna see a movie?", "Yah" stuff, then your concerns are misplaced.
(no subject)
Date: 2008-03-28 03:13 pm (UTC)Possibly not, and I don't think I'm going to support it *soon*. But this whole line of thinking started with thinking about how the system might be used for purposes of activism, and I can see activists in, say, China caring about that. (Remember that Facebook is just the initial platform -- I don't consider it a fundamental part of the system.) And as you suggest below, businesses certainly might care.
So I'm at least pondering the end-to-end encryption scenario. I think it's at least a year off, but I'm going to put it on the story list and let it stew. We'll see if it reaches the top.
I suppose you could encrypt logs and other persistent data, and that you could make a good argument for doing so.
Yeah, but the problem is that doing so *truly* securely (securely enough to keep the government out) is a nearly intractable problem. Doing so for the logs (which are mostly write-only) is merely moderately hard; really doing so for the actual conversation data (which needs to be read/write) is really difficult. (I wouldn't have even credited it as possible, although
If you were building something like the old eRoom tool, or other workplace tools, then I could see how end-to-end confidentiality would be critical. Do you have such a model as your target? Or is this more "Whatya doin'", "Nuthin", "Wanna see a movie?", "Yah" stuff, then your concerns are misplaced.
Unclear. I suspect the initial target is a bit in-between those, just as LJ is -- that is, not completely trivial, but usually not earth-shaking. But business use is a very real and plausible long-term possibility that I have on the back burner. Frankly, I think that if this thing really took off in the popular sphere, people *would* start using it for business, just as they've picked up IM (there are obvious use cases, and business is getting more comfortable with hosted services), and providing additional capabilities for workplace is a very real possibility.
So we'll see. As with most of these upfront discussions, I'm mostly organizing my thoughts at this point. It's going to be a long time before this stuff matters, but I like exploring the problem space a ways in advance. (If nothing else, it tends to make me think of new possibilities of where the product might eventually go.)
(no subject)
Date: 2008-03-27 02:52 pm (UTC)What you can do that would make *me* feel better:
- when the owner of a conversation deletes it, everything should go away except a log entry saying 'owner deleted conversation'.
- logs should be minimal. You can't turn over information you don't have.
(no subject)
Date: 2008-03-27 03:29 pm (UTC)- To only accomodate subpoenas and such made via the US legal system. (ie: if a representative of some foreign government calls you up and tries to badger you into getting some conversation's info, the response would be "I'm sorry, no".)
- To explicitly say "while we don't provide tools for it, if you want to encrypt your conversation, feel free"; perhaps providing pointers to some simple tools therefor.
- To provide support for some sort of client-side encryption integration.
- For those "by word-of-mouth invite only" conversations, support encryption of whatever conversation data is stored in the DB. I.e.: the actual passphrase for the conversation is never stored on your servers [just a one-way hash of it for "you can view this" authentication], and is used as a key for some form of encryption pre-storage and decryption post-retrieve. This is, admittedly, of limited benefit - it wouldn't help if you became legally required to capture the information on an encrypted conversation, since it would be trivial to change your code such that you stored that incoming key the next time someone used it. But it would mean that no-longer-active conversations of that sort couldn't be accessed (since nobody would be sending the key over anymore). It would also provide a little bit of extra protection vs. hackers (regular or state-sponsored); someone simply grabbing a copy of your database wouldn't be able to access those conversations without also doing the above sort of monitoring.
(no subject)
Date: 2008-03-28 02:43 pm (UTC)Good point -- I'd pretty much been intending that anyway. Indeed, one of the reasons I'm thinking about keeping the company closely held is specifically so I can tell China to buzz off if need be. It's hard for a public company to sacrifice that large a market (it's the thin end of the wedge for Google in terms of evil, far as I can tell), and I'd like the ability to do so if principle requires.
To explicitly say "while we don't provide tools for it, if you want to encrypt your conversation, feel free"; perhaps providing pointers to some simple tools therefor.
To provide support for some sort of client-side encryption integration.
Interesting. I hadn't thought about that, and it has some important implications, but it's worth thinking about. (In particular, it implies that there needs to be at least API access that provides a completely unmodified data stream, which is not otherwise obvious given data sanitization issues.) I may need to add that one to the story list, because it's a neat and useful idea.
For those "by word-of-mouth invite only" conversations, support encryption of whatever conversation data is stored in the DB.
An interesting middle-ground suggestion. Hadn't occurred to me, precisely because it's a fairly weak protection, but you're right that it has some benefits. Worth considering, anyway -- thanks!
(no subject)
Date: 2008-03-28 02:34 pm (UTC)While technically true, it's rather oversimplified. Privacy is a spectrum, not a simple black-and-white. It's all about who you're trusting, for what, and how much. The issue for me is figuring out what I'm going to provide in that regard, and setting expectations appropriately.
when the owner of a conversation deletes it, everything should go away except a log entry saying 'owner deleted conversation'.
Reasonable, but plays very closely into the ownership question. I still haven't decided how I'm going to resolve that particular tension. (Deleting conversations is a medium-high story, but still a ways off yet in absolute calendar time.)
logs should be minimal. You can't turn over information you don't have.
A reasonable point. Fortunately, also probably a moot one: I probably just can't *afford* extensive logging -- it tends to be far more trouble than it's worth once a system is past initial debugging...
(no subject)
Date: 2008-03-27 05:15 pm (UTC)(no subject)
Date: 2008-03-28 02:44 pm (UTC)(no subject)
Date: 2008-03-27 05:17 pm (UTC)If you want to encrypt private conversations such that only legitimate readers have decryption keys, more power to you, and you should advertise as such (eg "if you lose the key, the data is gone, and we have no way to recover it").
(no subject)
Date: 2008-03-28 02:48 pm (UTC)Sad, but a very good point. It's going to be some time, at least, before the company is big enough to run its own servers.
If you want to encrypt private conversations such that only legitimate readers have decryption keys, more power to you, and you should advertise as such (eg "if you lose the key, the data is gone, and we have no way to recover it").
This would be lovely in principle; unfortunately, I don't think the mass-market web has the necessary infrastructure yet to do this easily.
That said, it's worth thinking about how that could be encouraged. For instance, I could easily see working with Firefox plugins to support genuine end-to-end security. That's a long-term story (that is, not in the next year), but potentially a very interesting one, and there's nothing technically *hard* about it -- it's just a lot of work...
(no subject)
Date: 2008-03-28 12:04 am (UTC)Now if I understand things correctly, once the government comes calling, you aren't allowed to go telling your customers to run for cover. However, if they care, they can see it for themselves, once you are forced over into switching into some other way to do the key checks, and then they can make what inferences they would.
Possible...
Date: 2008-03-28 12:59 pm (UTC)Of course, the problem then is that you're exporting crypto software, which is still regulated.
Re: Possible...
Date: 2008-03-28 02:26 pm (UTC)Re: Possible...
Date: 2008-03-28 02:58 pm (UTC)Of course, all the browsers include public-key crypto, to implement SSL; in theory, they could expose that to JavaScript. Um...<dig, dig>...Mozilla does that, but it looks like it might not let JavaScript encrypt anything; it might be just for signing.
There are various all-JavaScript implementations of cryto code, including at least one public-key implementation.
Re: Possible...
Date: 2008-03-29 03:57 pm (UTC)Hmm. Honestly hadn't occurred to me -- I wonder how slow it is? The problem at that point is just key management, which is still an issue, since the idea of good browser-side storage is still kind of new and immature. Might be more plausible in a year or so, though...
Re: Possible...
Date: 2008-03-30 01:23 am (UTC)(no subject)
Date: 2008-03-28 02:56 pm (UTC)It's a complex enough problem that I can't just support it upfront, but it's easy for me to mentally sketch the architecture (provided the users have a pluggable browser such as Firefox), and it would provide the users with near-optimal security if they care. (And it's still a good deal easier to implement than trying to manage keys across borders, and means that I am not actually involved in the cryptography itself, with all of its legal complications.)