Friday, 5 September 2014

On the time of whom?

Another Sunday morning Speculation, (where one of the many things that puzzle me get posted in the form of a research hypothesis.)

It is the start of a new academic year for many and that gives me the perfect setting for something that I've wondered for a while:

Imagine that ten women arrive at University and are all put together in the same house. They have communal everything and sleep in a long dormitory with five beds down one side and five down the other, (more like a boot-camp than a University.)

My question, (and I can find no papers on this) is this: presuming that when they arrive they are not all menstruating on the same day of the month, and that with time they sync up - which girl is the anchor that they move into sync with, (if any - maybe it is more egalitarian than patriarchal and they all move the required time to arrive at a mean time, (statistics, not puns!) )

Then the next questions are: how and why? If it is controlled by pheromones then it should be possible to create a contraceptive mist that would change cyclically to avoid the more rapid sections of hormonal fluctuation that exist, (e.g. day 13 oestrogen crash). Which could lead to happier lives, (but further speculation is poetic at best and futile or damaging, until the first question is confirmed.)

Monday, 4 August 2014

Dark thinking with Quaternions

When I hear about the fourth, (or more) spacial dimensions it reminds me of
Quaternion multiplication.

I am impressed the indignation for the square-root of -1 seemed to be an impediment to an equation, and the gall of using i as a solution. I'm amused that in many cases that provides the solution to an actual equation resulting in a solution back in the plain of the number line.

Just as we can avoid the physical limits of three gimbals and the associated problems with Euler angles by converting our calculation to using quaternions I would argue that i,j and k do not exist, but that we need to imagine that they do in order for our minds to accept the logic that teleports us from problem to solution.

Friday, 27 June 2014

Dark Energy vs Dark Matter

(some) Speculation, (not proven science!)

Dark Matter = missing mass of a galaxy, (about 26.8% of stuff)
Dark Energy = fuel the expansion of the universe (about 68.3% of stuff)

If all visible matter from the standard model is only about 5% of mass-energy within the space-time of this universe, and theoretically, about half of all mass-energy should be anti-matter, then 13.4% of stuff should be anti-dark matter and 34.15% should be anti-dark energy.

Until we prove otherwise I'm going to presume that Dark Matter is cosmic phlogiston and the measured variations in the rotation of stars within some galaxies is caused by a property of either space-time or the interaction between space-time and mass-energy.

Dark Energy, on the other hand is proof of other {mem}brains interacting with this one.

Bonus round

Is there an experiment to demonstrate the speed of force through a solid? (To disprove the "fastest way to communicate with the moon is to send Morse code by constructing a solid rod all the way from earth, and pushing on one the Earth end to press the Morse key on the moon.") I can see that the inertia of the rod would prevent the message getting there faster than a laser photon, but how much slower would it be?

Wednesday, 25 June 2014

Programming paradigms and economic modles

(Haskell has a functional paradigm while Objective-C is imperative.)

"If  John Maynard Keynes economic model can be thought of as being functional then Friedrich Hayek's model would be imperative."

Discuss. Which is more socialist.

Keynes seems to see the macro financial system as being as static as a Haskell variable, which leads to the conclusion that a government can inject/invest its way out of a financial crisis. This idea comes from the perceived perpetual existence of society and that a nation is a perpetual-money-machine that sometimes needs a kick.

F.A. Hayek focuses on the individual micro economics and scales up, (ad absurdum from Keyne's point of view.) This is more realistic because indefinite growth on a physically finite planet is ludicrus.

Keynes might point out that money, like numbers themselves are infinite.

Tuesday, 24 June 2014

Stairing at a Designing

When I first saw MinuteLabs Corner-reflector I was designing a new wooden staircase. It was to connect the upper floor of a house with a lower one, but with some unusual constrains. I looked into the local laws that I could find, and the usual calculations dictated by logic and human proportions, (I didn't find Stairs Calculator until much later.)

To that end I plan to create an interactive system where by you enter the height, [0] (between the ground level of the upper floor and the lower level) and the depth within which the staircase has to fit. You can toggle risers on or off, (for steep stairs they would be off automatically.)

Then the system would calculate how many steps you would, (or should) need and how close to being a ladder, (vertical) or decking, (horizontal).

The BlockLayer site even calculates the piece of wood that you need. The only option that it seems to be missing is the ability to set the treads into slots in the stringer, (the long side part of the staircase) rather than notching.

Once I dig out my notes, I'll add the measurements, calculations and regulations that I found.

[0] Actually just dragging the staircase like the laser in the corner-reflector seems like a better human interaction design.

Sunday, 8 June 2014

Windows Desktop Automation

Have you ever needed to have your mouse click over and over while you were not there? Do you want to set up hotkeys to auto-complete a piece of text, (like an email sign-off) that you find yourself doing over and over each day?

I downloaded nine different piece of software and "Auto Clicker", though commercial, did seem to be winning.

I wanted something that could click three buttons, in order, with a 10 second pause between loops. Most importantly it had to be able to loop until I stopped it.

Then I found AutoHotKey. It was not the easiest to use. I did have to read the amazingly well written manual, (and in this day-n-age it seems that no one wants to read manuals, or rather RTFM==design_flaw.)

It reminded me of NSIS (Nullsoft Scriptable Install System) and one I got used to it, (about five minutes) I started to see just how powerful AutoHotKey is. First of all I added:

F6:: Reload ; why isn't this in the default script?

So that I could reload the changes as I hacked them. You see AHK is powered by a simple, yet hugely comprehensive, scripting language.

But wait - there's more! I was thinking, "wouldn't it be nice if I could train AHK." I wanted to be able to do the clicking on the screen and have it just learn. I didn't want to have to use the "Window Spy" that came with it. Well with v1.0.48.05 there is, ( C:\Program Files (x86)\AutoHotKey\AutoScriptWriter\AutoScriptWriter.exe which will do just that!.

So if you have written a program, (or even just a bash script) then AHK will be simple. If email confounds you then 's "Auto Clicker Typer" might be more your speed, (though it would not loop for ever, I could get close with ACT by using copy-n-paste to duplicate my three buttons many times, (a few hundred) and then setting the Loop option to over nine thousand.) The biggest problem with ACT was the errors when it tried to access the HKEY_LOCAL_MACHINE\Software\AutoClickerTyper\AutoScript entry in the registry, (though clicking Continue seemed to work in most cases.)

So AutoHotKey is officially added to my list of "good stuff" in the windows software category. Thank you Chris Mallet and others.

Tuesday, 27 May 2014


We are each surrounded by information. This data is one way to store, catalogue and process all aspects of our lives. The miasma of data does not belong to us, but the parts that we want and need to store and access certainly can feel very personal. I don't share my pa$$word with anyone. If they have to have access they are issued with their own key.

For the past couple of years, in my spare time, I have been thinking about the most practical ways to store this data.

Then I met tiddlywiki. The moment that I saw how tiddlywiki could work with a server I saw one way to unify all of your personal data; securely store it in a distributed fashion and give others access. Even the configuration was going to simply be another piece of data within miasma, (or rather me-as-data.)

Each atom of information is a "ma" - deliberately still dividable, (literally into two separate letters, but metaphorically this means data systems can divide a piece of data into arbitrary blocks and encrypt and store them as efficiently as possible. (Even extract and re-encrypt if a better or more secure system is discovered.)

Using Shamir's Secret Sharing access can be granted, or revoked to the granularity required by the user.

I wanted the system to be able to store itself on multiple locations and have the flexibility to automatically merge distributed collections between locations and time, (credit to git-scm.)

So each ma of data has to have the option of being encrypted in one or more ways, depending upon the access of the recipient. For example, we can sign the data with one of our secret keys and encrypt the data with the public key of the recipient. For a transitory message, "shall we have pizza tonight?" that is not only over kill, it is disempowering. Communities may agree that certain individuals should have access to that piece of information. [0] Through meta-data mining a corporation could profit from sending your pizza vouchers or some other action, when clearly your message was not meant for public consumption. This is passive corporate abuse, (I'm working hard not to call it digital-rape.) Just because person A's perception of person B is interpreted as communicating one intention [1], it does NOT enter person B into any form of contract with person A and it does not entitle person A to use person B or to push them into having to respond. (Yes, I'm tired of spam, cold-calls and heckles.) There are too many people in the world for society to demand a "no" when a lack of response should equate to "leave me alone, I don't even want to burn through another second of my short existence receiving an apology from you." (Though large cash compensation is fine, as long as you don't imagine that you are purchasing anything. Think of giving money as a modern form of plenary indulgence.)

So as you can see - the moment that you move out of the field of math and into data you have to consider the wider social aspects, (and they are many and varied.)

The Practical


I envisage a tiddlywiki style javascript overlay that can be loaded from any hosting site, ( github, pastebin, your local USB flashdrive) from a bookmark as an overlay to any, (and all, or just HTML5 Local Storage [2]) data storage, (webmail). That can leverage the remote site as a data storage system, (unsubscribed IMAP tree) to store the various pieces of data encrypted with a symetrical key, (AES) and re-encryptable on the fly, (decrypt -> encrypt{3DES} -> store) the moment that the existing key is compromised or the protocol falls to cryptoanalysis.

The system, (lets call is nuage) will also be able to share collections of data with individuals, (if it is an overlay to a webmail, then it can 7zip the data into an attachment and send it, and if they recipient has the same system, then it can process the message without user intervention.)

Most importantly the system will let you dump a backup to local disk as an encrypted archive, (either encrypted with a passphrase that you chose or with one that you never know that rests within the system's own meta-metadata and is shared to one of your nuage remote locations.)

Lastly - it has to be distributed. A copy can be forked onto or into any and all storage media and self merging. (Yes, that's going to be fun for someone.)

So each "ma" could take the form of


Though it is important that the meta data of ver(sion) cdate, adate, mdate, be optional and that the type of encryption be stated either at the start of the string, (thanks dovecot) or as a separate meta string "enc-type":"elgamal" with the option of the "key ID":"EE87B9F78EFC079FB0A4B4DE4AC889DAFA479164" (there will be times when linking a piece of data with a key will be damaging, so having the ability to leave the user to store that piece of information in their head is important. That said, I've found that addresses and especially passwords should have a data on them, so that the system can let me know which ones have not been changed for $update_pw_after_x_days.)

At other times you want to be able to identify a piece of data so:

(Obviously this would also have to reference the public key in question.)

There is a lot of room for expansion, and thankfully javascript has matured, (and more importantly the libraries now exist) to be able to implement this idea.

[0] police access to encryption keys
[1] "We are both here for the same thing!"
[2] due to the limits of locl storage, a progress bar must be available.

Keep your enimies classy

It has been said that you can tell more about a person by their enemies than their friends, (and if it hasn't - I just did.)

I want you to think about two words: Viking Philistines.

The Northmen that attacked and invaded the north east of England were from warrior nations. They were not gentle or fluffy and didn't have funny hats with horns. They were brutal and eventually assimilated or repelled.
(While in Normandy they were more successful and the Frankish people built castles to keep them from spreading.)

The Philistines were a tribe, as labelled and described by another tribe. Among the many Philistine warriors there is one that you have certainly heard of, (though some need the addition of his vanquisher to give him context): Goliath, of "David and Goliath" from a number of religious stories.

So what does that say about how the attacked tribes and those that record these events view and treat these historic tribes.

For me Viking is a complement, (almost romantic word - certainly when it comes to Lagertha Lothbrok). Philistine is an insult.

Possibly more interesting: How did these few words make you react emotionally? Was it interesting, antagonistic, frustrating, enlightening? How much control over your own reaction did you have? If you had been born in a different country, would you still have a similar reaction? Or is your community dissociated from the country within which you live? I bring this up based on the conflicting ideas of:

You live in your body;
You ARE your body.

Tuesday, 18 March 2014

[android-app] hiker-zen


Citizen hitch-hiker app: The app works with the google maps app to locate people on your path that share some or all of your journey.

How it works

The driver, (CHD) searches for their destination, (google maps called from hiker-zen) and indicates which vehicle they are using, (car, van, bus) and any variation, (one back seat is taken with a child.)

The app locates citizen-passenger-hikers, (CPH) that are located along the path, (the app 'rounds' your location to the nearest bus stop or other feature so that it does not leak your home-address.)

Preferences can be set by the driver, (height, weight, gender, age) for the CPH. (Or just people in your contact list or social-network.)

When the two mobile devices are brought into proximity the CHD and CPH pair their devices, (blue-tooth, wifi, Infrared). This then starts to tally the social capital, (two citizen-hiker currencies) of distance and time. (You may take someone quickly on a long motorway or slowly through inner-city traffic.)

The capitol can be used to verify a user, (dynamic web of trust) to suggest remuneration, (shared cost of shipowner or fuel costs) and eventually in other locations like parking fees.

Passengers can "fine" a driver for:
agressive driving;
Inconsiderate driving;
dangerous driving;
failing to indicate at a roundabout or junction.

The simplified interface would simply have a thumbs down for "I feel uncomfortable" and a thumbs up for "I'm impressed", when the CPD makes space to let another driver into the flow of traffic.

After the journey the users can vote, "I would do it again" for that journey with that, (those) people. If it is reciprocated then that connection would be weighted preferentially, (though there must be a 'break-up' function so that the user can manually veto a user at a later date, and 'make-up' to remove the veto.)

Abuse concerns

How do we prevent abduction?

  The past may be no indication of the future but as each bit of social capitol is part of a digital crypto-currency and each event is signed by the users instance of the app, users should be able to determine compatibility. Another attack would be for the app to automatically 'tweet' the other persons unique ID, and location - so the two citizens don't commence the shared portion of their journey until the public notification is acknowledged by the public forum/email server.

How do we prevent mining?

The two mobile devices should be able to verify that they are on the same journey from sensor data, (GPS, accelerometer, barometer). The real question is how do we digitally encode a users data into their app so that a malicious person can't simple purchase, (steal) ten devices, invent a bus and then drive the devices about on the back of a bicycle [0] to 'mine' social capitol? (Capitol that could later be used to fraudulently impersonate a good citizen with the intent to perform an unacceptable act.

Spot checks by curtain twitchers: A twitcher (CTW) could be notified of an alert [1]

If the CTW thinks that they have seen the event they can verify, (by pressing the confirm button) or by taking a picture.

Other consers?
Do you see any problems with this idea and how to solve them? How would you implement this?

Some versions already exist for real-time car-pooling.

[0] No reason why a CHD can't give give a lift to a CPH on a bicycle, but it would be unusual for a CHD to regularly lift ten CPH at the same time.
[1] Probably in JSON: {journey: { vehiclue: {type: "car", colour:"blue", make: "Bently"};  people: [adult: "2", child: "3"] };

The problem with perception is that people: [ men: "2", women: "2" ] is open to interpretation of gender expression, so the app will have to be flexible and as granular as the CHD and CPH want to express themselves that day. The app will have to take the simplest common denominator, so if a vehicle has {people: {men: "1", women: "3", adult: "2" } } then the app for the CTW would display, "6 adults?" as the question.
Infants are often not visible but the app should have them and pets as categories.

As with the people the vehicle details will have to be simplified and obfuscated, "A red sedan with license 41H ***?"

Wednesday, 5 March 2014

Tai Chi all of the muscles groups?

 Whether you divide the body into 8: {shoulders, arms, back, abdomen, chest, butt, thighs and calves.},  11: { arm:{fore,bi,tri,shoulder} core: {chest,trap,lat'dorci,abs} leg: {quad,ham,calf} } or more major muscle groups, we can probably agree that we use some muscles to exercise.

Mentally, as I exercise, I observe internally, (if I pay attention):
  • face: eyes, jaw
  • neck
  • upper back and shoulders
  • biceps
  • triceps
  • lower back
  • lats (Obliques)
  • core
  • gluteus
  • hands/forearm
  • thigh (anterior extensor)
  • medial adductor (inner-thigh)
  • hamstring (posterior flexor)
  • calves
and eight areas of articulation: ankles, knees, hips, waist, shoulders, elbows, wrists, neck.
Each of these 'groups' are used in one or more of the sports that I have actively participated in. Without getting distracted by my incorrect list lets assume that it is correct and look at the Beijing 24/ Yang style short-form Tai chi. Specifically what does each movement focus upon physically, (other than balance and breathing.)

[Make your list and then notice which groups are worked upon the least.]

This is obviously the wrong way to look at Tai Chi or any 'Eastern medicine' as it is intended to be holistic. (Each movement may work on multiple major groups, and subtly upon many of them.) But if you take the Yang style long form and reduce it to 24 movements, are you sacrificing one of the muscle groups in my list?

Why does this matter? Well I was doing some exercise research that worked on my core and then did my Tai Chi, (short-form.) I realised that only in "kick with the heel" did I flex my sore lower abdomen.

This made me go through the eight pieces of brocade to check muscle groups and articulation points and found that it seemed to cover more groups in fewer movements. Also I could engage my abs or not in some of the movements. Hardly conclusive, but if Tai Chi were to be a supplement to the 8PB, which Tai Chi movements would have the least overlap with those eight movements?

Thursday, 23 January 2014

How deep is your FUD?

When as executive from Microsoft said, "You can just google me, ah um Bing me." There was much chuckling. At the time it was said that Bing was not making money and it was suggested in some publications that they would be happy for someone else to take it off their hands.

Was this a natural slip of the tongue?

If you don't have enough material for your conspiracy theories then try this idea on for size:

"It was a corporate equivalent of a military surgical strike. If Google becomes google in law then their trademark becomes worthless."

Just a moment? What are you on about? (Would be a reasonable question.)

Well, (I would say with deliberate pause) if it can be shown that a trademark has become part of the normal term of a man-in-the-street then it is watered down like a sandcastle on a beach. Just as band-aid, in the USA become the default term for a self-adhesive medical strip and  Hoover, (the vacuum cleaner manufacturer) became the standard word for that item in the UK, if google replaces the word search on the Web and for those that use the Internet, then Google will have hit that interesting point in law that creates a sort of they-should-have-already-made-enough-money point in its corporate life cycle.

If you want to season that plate with some cynicism, (and it is Monday after all) are mistakes and apologies the current state-of-the-art in the corporate PR arsenal?

Monday, 13 January 2014

Outlook 2013 vs Exim 4

I had problems getting Outlook 2013 to play TLS with my exim servers. The solution seemed to be to set Outlook's encryption to 'Auto' and use port 587 (a NON TLS port).  I have both PLAIN and LOGIN auth but Outlook uses the latter. If you are still using a flat auth file that uses the CRYPT hash then there is an example line for you, but I mostly authenticate against a database,
(so that changes don't have to be rolled out in batches.)

exim config snippit:

tls_advertise_hosts = *
daemon_smtp_ports = 25 : 465 : 587
tls_on_connect_ports = 465 : 6465
# some ISPs filter 25 and 465 to their own SMTP servers for 'simplicity' hence I have  6465 for customers with that affliction.

MYSQL_AUTHPLAIN=SELECT  im_server FROM imap,domains WHERE imap.im_doid = domains.do_id and concat(imap.im_userid,'@',domains.do_name) = '$2' ) AND ( im_auth='${hmac{md5}{$3}{$3}}' || im_auth=encrypt('$3',im_auth) || im_auth='${sha1:$3}' )
# transitioning from encrypt to sha1 and merging in an hmac_md5 config

MYSQL_AUTHLOGIN=SELECT  im_server FROM imap,domains WHERE imap.im_doid = domains.do_id and concat(imap.im_userid,'@',domains.do_name) = '$1' AND (  im_auth=encrypt('$2',im_auth) || im_auth='${sha1:$2}' )

begin authenticators
# $1 is the old string for $auth1; $2 = $auth2; $auth3 = $3

  driver = plaintext
  public_name = PLAIN
  server_condition = ${lookup mysql{MYSQL_AUTHPLAIN}{1}fail}
  server_advertise_condition = ${if def:tls_cipher }
  server_set_id = $2
 driver = plaintext
 public_name = LOGIN
 server_prompts = "Username:: : Password::"
 #  server_condition = ${lookup mysql{MYSQL_AUTHLOGIN}{1}fail}
 server_condition = "${if crypteq{$auth2}{${extract{1}{:}{${lookup{$auth1}lsearch{/etc/exim/passwd}{$value}{*:*}}}}}{1}{0}}"
 server_advertise_condition = ${if eq{$tls_cipher}{}{no}{yes}}
 server_set_id = $1


Example mysql schema, with domains in one table and imap, (and smtp authentication in im_auth) in another:

CREATE TABLE `domains` (
  `do_id` int(255) NOT NULL AUTO_INCREMENT,
  `do_name` varchar(255) NOT NULL,
  `do_status` enum('disabled','suspended','enabled','migrating out','migrating in','registering','desired','disputed','remote') NOT NULL DEFAULT 'remote',
  `do_added` datetime NOT NULL,
  `do_acid` int(255) NOT NULL COMMENT "account id - other table",
  `do_group` int(255) DEFAULT NULL,
  `do_peid` int(255) DEFAULT '0' COMMENT "people ID",
  `do_location` varchar(255) DEFAULT NULL,
  `do_masters` varchar(255) DEFAULT NULL COMMENT 'a ; delimited list of ip addresses',
  PRIMARY KEY (`do_id`)

  `im_id` int(255) NOT NULL AUTO_INCREMENT,
  `im_userid` varchar(128) NOT NULL COMMENT 'the bit before the at sign',
  `im_doid` int(255) NOT NULL COMMENT 'link to domains.do_id',
  `im_passwd` varchar(74) DEFAULT NULL COMMENT '{HASH}string e.g. {SHA1}shy75adsgf=',
  `im_home` varchar(255) NOT NULL COMMENT 'explicit path on im_server',
  `im_uid` int(11) NOT NULL COMMENT 'probably 8 (mail) though for shell users set it to their uid',
  `im_gid` int(11) NOT NULL COMMENT 'probably 12 (mail) or 8 on some systems',
  `im_server` varchar(128) DEFAULT NULL COMMENT 'mostly this will be the localhost or hostname',
  `im_quota` int(255) DEFAULT NULL COMMENT 'In Megs: 2 petabyte limit',
  `im_peid` int(255) DEFAULT NULL COMMENT 'links to people table',
  `im_auth` varchar(255) DEFAULT NULL COMMENT 'exim authenticates from this if it does not understand im_passwd - useful for migrating from MD5 to SHA256',
  `im_mode` char(4) DEFAULT '0640' COMMENT 'smallint seems wrong',
  `im_dir_mode` char(4) DEFAULT NULL COMMENT 'exim file and dir modes',
  `im_last_seen` datetime DEFAULT '0000-00-00 00:00:00' COMMENT 'the last SMTP,IMAP',
  PRIMARY KEY (`im_id`),
  UNIQUE KEY `im_row` (`im_userid`,`im_doid`)

# I've never had to add a NULL imap row to enable SMTP, but that is perfectly possible.

Tuesday, 7 January 2014

SFR is rubbish at IP networks

That title should get someone's attention. Why such a deliberately childish title? Because I could not easily find a "Network Status" page or a "report network problems" page. One would have been able to reassure me that the issue is being dealt with and the other would have let me, (someone that knows a little about IP networks) to provide useful information.

Talking of which: (notice the huge jump between hop 04 and 05.

03.        24.0%   146   27.2  57.2  24.6 640.1  88.5
04.                   25.5%   146   25.6  72.8  24.4 585.3 100.4
05.        26.9%   146  5608. 969.8  26.3 7250. 2008.
06.        30.3%   146  5768. 868.6  28.6 7180. 1863.
07.         28.3%   146  6026. 866.7  33.2 7364. 1866.
08.        26.4%   145  5782. 1018.  34.5 7118. 2045.
09. ix-28-0.tcore1.PVU-Paris.as6453 28.5%   145  5506. 726.2  32.2 7155. 1693.
10.                   29.2%   145  5745. 734.4  32.6 6676. 1685.

I'm on a friends ADSL line and this has been happening on and off for at least a week. I'll add more traceroutes each time I remember. I can say that the last time I was here the house was using Darty, (which used someone else's network) and that was much better. SFR seems to be good at mobile but crap at home networks, (from one data point between 2013 and the start of 2014).

UPDATE:  2014-01-16 After reporting the problem, (they seemed to be able to instantly fix it) and mentioning it to a cold-caller trying to sell anti-virus on behalf of SFR the core problem is still showing:

2.     31.1%   212   24.4  56.9  22.9 344.7  67.7
 3.        28.8%   212   25.8  51.0  22.4 288.1  52.7       30.2%   212   43.1  51.9  22.5 359.4  56.7
 5.        29.2%   212   28.2 2573.  23.5 6441. 2508.
 6.        29.2%   212   28.1 2696.  25.4 6866. 2512.
 7.          34.4%   212   35.3 2555.  31.6 6737. 2509.
 8.        34.4%   212  137.3 2659.  32.3 6719. 2508.
 9.  32.5%   212   85.3 2630.  44.6 6745. 2512.
10. 35.8%   212   38.5 2506.  30.2 6729. 2530.

and the same evening:

 Host                               Loss%   Snt   Last   Avg  Best  Wrst StDev
 1. box                              4.4%   479    2.3  39.3   0.8 1399. 132.4
 2.      28.9%   479   24.6  73.5  22.8 1360. 157.6
 3.        26.6%   479   26.1  72.1  22.7 1303. 150.6
 4.       25.9%   479   39.9  67.6  22.6 1253. 140.2
 5.        32.8%   479  5242. 2126.  23.2 10878 2510.
 6.        28.5%   479  5427. 2293.  25.3 11725 2617.
 7.         28.9%   479  5129. 2300.  30.3 11147 2569.
 8.        29.7%   479  5075. 2242.  31.6 12060 2648.
 9. ix-28-0.tcore1.PVU-Paris.as6453 31.0%   478  5412. 2312.  43.3 11581 2566.
10. if-12-2.tcore1.PYE-Paris.as6453 31.4%   478  4962. 2217.  30.2 13397 2593.

(Feel free to add your own traceroutes as I am still unable to locate where in their website you can report problems to them - OR even find network status.)

The Conclusion:

SFR happened to set one of their 'would you like to buy anti-virus' sales men on us - which didn't go as he had hoped, but was probably the most effective way to inject a complaint into their company. On top of that SFR were informed at each end of the week, (you can call them on 1023 in France or 0033 6 1000 1023 from the UK.)

The first time they were able to almost instantly drop the latency from ~4000ms to 32ms within their backbone. (Daily and sometimes hourly cold-reboots of the router in the house all week), lead to Thursday where the problem was reported again. A very frustrated, (and a little bit angry) tech support bloke rebooted the house router remotely, (killing the crackly phone conversation stone dead); though it did fix the problem - and since then it has been a lot better. Still not perfect but actually useable. So what was the problem? What can we guess. Is their backbone capacity too heavily over subscribed? (My first guess, but this isn't 1998.) No, I'm almost certain that it is entirely down to some complicated rate-limiting withing their network somewhere between hop 4 and 5:

4.   <= no visable problems from users end 
5.   <= 50 times as long for packets to return (if they bother.)

At the time of this problem I was in the habit of starting with a trace to one of google's public caching name servers. The problem here is that SFR seem to have a direct peering with google via an address in and probably have different rules for that connection, (after all SFR would just be shooting themselves in the head - as opposed to the other foot - if they slowed down their customers connection to google.)

About this blog

Sort of a test blog... until it isn't