4 ms·

They wouldn't be able to do this on the server side. However their native iOS, Mac, Windows clients could just create&update the search index locally and sync t

They wouldn't be able to do this on the server side. However their native iOS, Mac, Windows clients could just create&update the search index locally and sync the encrypted version. With HTML5 localStorage this also works in Web Apps. Of course there's a tradeoff, but for me that's easily worth it.

derefr 14y ago
The "tradeoff" seems to be "make the server into a dumb store for encrypted data." At which point, you don't have Evernote (an API for fuzzy-matching clippings punted into it from various devices), you have Evernote (a set of fat client programs each of which must maintain an entire copy of the dataset--notably, stored necessarily alongside its own decryption key on the client side, increasing attack surface--and do manual synchronization) plus a POSS (Plain Old Storage Service, like S3.) The resultant workflow sounds like it would have more in common with several people trying to edit the same Word document over SMB, than with making web requests. [EDIT: clarity about client-side encryption]
- MartinMond 14y ago
  The "tradeoff" seems to be "make the server into a dumb store for encrypted data." At which point, you don't have Evernote (an API for fuzzy-matching clippings punted into it from various devices), you have Evernote (a set of fat client programs) plus a POSS (Plain Old Storage Service, like S3.) In fact, the workflow sounds like it would have more in common with editing a Word document over SMB than with making web requests. Not at all. Fuzzy-matching can be done client side just as well as server-side. The constraints are a bit different, but not too much (E.g. on the server-side: Make it scale --> conserve CPU, on the (mobile) client-side: Make it fast --> conserve CPU) Otherwise, yeah most Web Apps are nothing more than editing stuff over the network and visualizing it differently. The #1 competitor or a SaaS isn't some other SaaS but rather Word/Excel: http://www.startupcfo.ca/2011/05/the-1-competitor-for-saas-vendors/ http://www.startupcfo.ca/2011/05/the-1-competitor-for-saas-v... Edit: Also there's no reason to believe the client-side store wouldn't be encrypted. That'd be exceedingly stupid.
  - derefr 14y ago
    I don't think you got quite what I'm trying to say--the whole point of Evernote is that all your data is "there", and the collated index for finding this or that is "there", and so any individual piece of data doesn't need to be on this device or that device. When you insert a new piece of data, all you need is that piece of data. You send it off to Evernote, and they stick it into your database. Then, later--still without any local state--you search the server to find out if anything you've sent in matches some arbitrarily-complex query--and if there are results, then you temporarily download [the newest version of] them from the server. The collation itself--not the storage, not the UX, but the ability to immediately put something into device X and then find it using device Y--is what you're paying them for. Without that, what you have is a OneNote notebook stored in your Dropbox. > Edit: Also there's no reason to believe the client-side store wouldn't be encrypted. That'd be exceedingly stupid. It's encrypted with a key that's stored on the client, which is the same as the server being encrypted with a key stored on the server: effectively about as secure as DRM (i.e. not.) To put it another way: presuming you have a motive to gain access to just my data--with this hypothetical service, if you steal my phone, you have my data, and the key to decrypt it. Or if you steal my laptop, or my desktop, or any other device the service is synced to. Or if you hack into them. All you need after that is the passphrase I (hopefully) set to unlock my encryption key--and for a single target, social engineering (or lead-pipe cryptanalysis) can get that right quick. Meanwhile, there is only one thing people can do to steal my Evernote data: hack into Evernote's servers. If you just want my data, that's a whole lot more effort than it's worth, compared to just palming my phone. [Now, if you want a bunch of random people's data, this is where using passphrase-locked + per-account-salted encryption-keys server-side is actually relevant to security. If it takes O(N) time to crack N accounts, there's much less incentive to do it than if it's O(1).]
    - moe 14y ago
      Nonsense. If your imaginary attacker can social engineer your local encryption password out of you, he can just as well engineer your Evernote password out of you. There is no difference. The real issue is that today someone who hacks the Evernote servers will gain access to all users notes. With end-to-end encryption they'd have to target each user individually.

derefr 14y ago

The "tradeoff" seems to be "make the server into a dumb store for encrypted data." At which point, you don't have Evernote (an API for fuzzy-matching clippings punted into it from various devices), you have Evernote (a set of fat client programs each of which must maintain an entire copy of the dataset--notably, stored necessarily alongside its own decryption key on the client side, increasing attack surface--and do manual synchronization) plus a POSS (Plain Old Storage Service, like S3.) The resultant workflow sounds like it would have more in common with several people trying to edit the same Word document over SMB, than with making web requests. [EDIT: clarity about client-side encryption]

MartinMond 14y ago
The "tradeoff" seems to be "make the server into a dumb store for encrypted data." At which point, you don't have Evernote (an API for fuzzy-matching clippings punted into it from various devices), you have Evernote (a set of fat client programs) plus a POSS (Plain Old Storage Service, like S3.) In fact, the workflow sounds like it would have more in common with editing a Word document over SMB than with making web requests. Not at all. Fuzzy-matching can be done client side just as well as server-side. The constraints are a bit different, but not too much (E.g. on the server-side: Make it scale --> conserve CPU, on the (mobile) client-side: Make it fast --> conserve CPU) Otherwise, yeah most Web Apps are nothing more than editing stuff over the network and visualizing it differently. The #1 competitor or a SaaS isn't some other SaaS but rather Word/Excel: http://www.startupcfo.ca/2011/05/the-1-competitor-for-saas-vendors/ http://www.startupcfo.ca/2011/05/the-1-competitor-for-saas-v... Edit: Also there's no reason to believe the client-side store wouldn't be encrypted. That'd be exceedingly stupid.
- derefr 14y ago
  I don't think you got quite what I'm trying to say--the whole point of Evernote is that all your data is "there", and the collated index for finding this or that is "there", and so any individual piece of data doesn't need to be on this device or that device. When you insert a new piece of data, all you need is that piece of data. You send it off to Evernote, and they stick it into your database. Then, later--still without any local state--you search the server to find out if anything you've sent in matches some arbitrarily-complex query--and if there are results, then you temporarily download [the newest version of] them from the server. The collation itself--not the storage, not the UX, but the ability to immediately put something into device X and then find it using device Y--is what you're paying them for. Without that, what you have is a OneNote notebook stored in your Dropbox. > Edit: Also there's no reason to believe the client-side store wouldn't be encrypted. That'd be exceedingly stupid. It's encrypted with a key that's stored on the client, which is the same as the server being encrypted with a key stored on the server: effectively about as secure as DRM (i.e. not.) To put it another way: presuming you have a motive to gain access to just my data--with this hypothetical service, if you steal my phone, you have my data, and the key to decrypt it. Or if you steal my laptop, or my desktop, or any other device the service is synced to. Or if you hack into them. All you need after that is the passphrase I (hopefully) set to unlock my encryption key--and for a single target, social engineering (or lead-pipe cryptanalysis) can get that right quick. Meanwhile, there is only one thing people can do to steal my Evernote data: hack into Evernote's servers. If you just want my data, that's a whole lot more effort than it's worth, compared to just palming my phone. [Now, if you want a bunch of random people's data, this is where using passphrase-locked + per-account-salted encryption-keys server-side is actually relevant to security. If it takes O(N) time to crack N accounts, there's much less incentive to do it than if it's O(1).]
  - moe 14y ago
    Nonsense. If your imaginary attacker can social engineer your local encryption password out of you, he can just as well engineer your Evernote password out of you. There is no difference. The real issue is that today someone who hacks the Evernote servers will gain access to all users notes. With end-to-end encryption they'd have to target each user individually.