Ahah, your story is starting to make more sense! Setting the value above 8 would set the 'enable agg TX' bit in this bitmask.
Side note: 11n_disable=8 as a way to enable an advanced wifi settings is not the most user-friendly option in the world. It looks like this was added as a patch to fix a compatibility issue with certain NICs[1]. Be careful setting it higher than 8 -- as that would set other flags in the bitmask, potentially turning off 802.11n entirely!
I've not heard of AMPDU before -- but it seems to be an 802.11n-specific protocol extension that allows you to send multiple MPDU blocks (i.e. frames) to the router at a time, rather than one at a time. This seems to trade off latency for error correction between a client and the router [2][3]. So by setting this flag, you're able to transmit multiple frames at a time, rather than retrying each frame individually until it goes through on a busy channel. This really does seem to make a difference, one report online suggests tx throughput goes from 20Mbit/s to 80Mbit/s when agg TX is enabled[4].
Apparently this was introduced in 2014 in order to patch a firmware issue on certain intel NICs. There's a whole conversation on the linux-wireless list about it[5], which seems to conclude that they couldn't reproduce the issue[6]:
> I have seen reports from many users and many devices - and unfortunately I have never been able to reproduce - so I can't know if it affects all the devices. All I can see is that I saw reports on very old devices (5150) and with more recent devices (2230).
"mc" stands for midnight commander, which is itself a clone of Norton commander, which was sold by Symantec in the 80s and 90s. Norton commander was popular enough that by the time it was discontinued it had been ported to many other platforms (midnight commander being one of the ports for unix systems). Over time, Norton commander became the prototype for the "orthodox file manager", the category of file managers with two side-by-side file browser windows: many of the same key bindings for this category of file manager are the same as the original norton commander keybindings.
Midnight commander (and more generally an orthodox file manager) is especially useful for moving files from one system to another, as it usually supports browsing to remote file systems over ssh or ftp, and gives you a simultaneous view of where you're moving files from and to on the same screen.
> One of the unique design features of the microscope is the use of inexpensive spherical lenses rather than the precision-ground curved glass lenses used in traditional microscopes. These poppy-seed-sized lenses where originally mass produced in various sizes as an abrasive grit that was thrown into industrial tumblers to knock the rough edges off metal parts. In the simplest configuration of the Foldscope, one 17-cent lens is press-fit into a small hole in the center of the slide-mounting platform. Some of his more sophisticated versions use multiple lenses and filters.
It looks like it's a url-safe base64 encoding of an encrypted blob (perhaps a stack trace). They're using '-', and '_' instead of '/' and '+'. Replace these characters and base64 will succeed, leaving with ~3k bytes randomly distributed between 0x00 and 0xFF.
Looks like Google's reporting the issue resolved as of 12:06 PM PST. Our backends are working again, so I believe them :).
If you run things on App Engine, I highly recommend subscribing to the downtime notify list -- it's much more responsive and accurate than the App Engine status page in my experience.
App Engine's reliability massively improved with the HR datastore, and has gotten even better since the pricing change / SLA guarantee. It's actually remarkably good now, I recommend taking another look.
I wonder how much of Google's reliability comes from service failover and redundancy and how much from very reliable datacenters etc. I'd find it hard to believe that their platform and DCs aren't better than what we have seen from AWS, which could make compute engine a very attractive product.
A lot of it comes from automatic failover. Individual data centers have issues frequently, but if you're using High Replication Datastore then you won't notice it much, apart from occasionally seeing all your instances getting killed and restarted in a new datacenter, which also results in memcache getting reset.
"But that's the great thing about going this fast, the novel starts to eat you and you find yourself writing any time you can just for the pure joy of it. Even better, on the days where I broke 10k, I was also pulling fantastic words-per-hour numbers, 1600 - 2000 words per hour as opposed to my usual 1500. It was clear these days were special, but I didn't know how."
This thrill of creation, when you've got the design (momentarily?) cracked and all that's left is the small matter of putting to disk what's there in your mind already, describes some of the most exciting experiences I've had writing code.
I have a strange reaction to my epiphanies. When I've cracked the puzzle mentally and it's time to put the code in the text editor, I start getting extremely anxious. No clue why.
Edit: To clarify, I mean anxious associated with "anxiety", not "with great anticipation"
I know the feeling, and the reason why I start feeling anxious in the same way you do is because I know that afterwards I am going to be in a lull again, that I will be bored to tears and will have the hardest time concentrating and getting done what needs to be done until there is another hard part that needs to get done at which point I am flying again...
