1. 1

    I actually prefer this explanation; http://web.mit.edu/Kerberos/www/dialogue.html, perhaps its helpful to others as well

    1. 18

      Interesting read, with a few neat ideas (I liked the idea of local data caching).

      Why no ECC RAM though? The Xeon supports it so it would be almost silly not to use it. Oh, and a SAS LTO-4 tape drive won’t cost much more than a SCSI LTO-3 drive but it’ll hold twice as much and almost certainly be faster.

      1. 6

        Thanks for the tips. I didn’t know about ECC ram. Is memory corruption pretty common?

        1. 13

          Good question. Jeff Atwood discussed exactly this issue in 2015 and Dan Luu followed up with some further discussion. I don’t think it’s absolutely necessary, but if your CPU supports it, why not?

          Personally, I use it in every system I have that supports it. Yes, it’s more expensive, but why bother using something like ZFS (which I do) if you have no guarantee that bits aren’t getting flipped before they even reach the disk?

          1. 10

            Jeff Atwood cites google as the case for not using ECC.

            Emulating google is also emulating googles mistakes.

            There is some research to suggest memory corruption occurs especially at 8+ GB scales. Perhaps these problems scale with size so at 64GB it is likelier.

            ECC in a personal machine is a trivial cost. At google scale even trivial cost can matter, but in this case, unless this is a gaming toy use ECC. I mean why not?

            Some HN discution on the topic

            https://news.ycombinator.com/item?id=14206811

            1. 6

              From that discussion:

              While I was at Google, someone asked one of the very early Googlers (I think it was Craig Silverstein, but it may’ve been Jeff Dean) what was the biggest mistake in their Google career, and they said “Not using ECC memory on early servers.”

          2. 2

            Another benefit of tape you should’ve mentioned is increased longevity and better recovery. Many cheap mediums have worse longevity than they advertise. DVD-R’s, for instance, can become unreadable in mere years. Enterprise focus on future proofing with lots of tape use means it’s unlikely to disappear like Zip or MO’s. Less uncertaintly than things like BluRay.

            The only remaining comparison are to cheap RAID arrays. Idk where they are in price per GB right now.

            1. 2

              Great point, I’m drafting an update to the article and will include this.

        1. 1

          Well, while interesting, this requires users to install an application.

          I find it somewhat more worrisome when information inserted into the dailer can be compromised through side channels such as sensors, which in the case of android requires no authorization to listen to and can not be turned of or feed invalid data. As such it seems trivial to capture pins and user ids that some organizations employ.

          https://blogs.ncl.ac.uk/security/2016/02/05/touchsignatures-identification-of-user-touch-actions-and-pins-based-on-mobile-sensor-data-via-javascript/

            1. 1

              It will be very interesting to see how the enterprise vs consumer drives compare next year.

              1. 1

                If I were AMD or any Intel competitor there is a opportunity competing not just on performance but on trust and ownership. I if I were them I would try to open up their equivalents like AMD Platform Security Processor and enjoy the more savvy crowd recommending their systems.

                1. 5

                  Well, as much as I hate to say it: this was only a matter of time, seeing how this was addressed at one of the more recent CCC conventions and pretty much ignored by telecom providers.

                  1. 3

                    Indeed here are the ccc videos from the other thread

                    https://lobste.rs/s/fb0lqr/after_years_warnings_mobile_network#c_bwf3w2

                    As I said this is not the only example of insecure networks that put too much trust in the other network actors; Telecom, payment processing at point of sale, travel bookings all do this. In the end putting the ordinary users at risk.

                    While sad for the victims, maybe this turns the tide on the banning encryption debates in parts of Europe.

                  1. 2

                    Here is the talk from Engel https://media.ccc.de/v/31c3_-_6249_-_en_-_saal_1_-_201412271715_-_ss7_locate_track_manipulate_-_tobias_engel it is really interesting.

                    “Private” networks that do not employ encryption and validation needs to stop.

                    Here is another example with the travel agencies networks https://media.ccc.de/v/33c3-7964-where_in_the_world_is_carmen_sandiego

                    1. 1

                      I find I tend to use !:x as in !:1 !:2 and so on instead of !$.

                      !:x takes the x parameter from the previous command, so from “mv /test /nexttest” !:1 is /test.

                      1. 1

                        a) this looks really cool!

                        b) python 2.7 or 3.x?

                        c) “utilize the benefits of multi-threading with minimal concern about the implementation details.” http://i.imgur.com/2RRXCb7.gif?1

                        1. 1

                          This wraps concurrent.futures which were introduced in python 3.2. Its less than 50 lines of code, and a good way to learn decorators, I recomend reading the source.