Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?

Comment: Use cases for http (Score 1) 319

by buchanmilne (#49599311) Attached to: Mozilla Begins To Move Towards HTTPS-Only Web

It seems Mozilla wants to move away from http, but here are some use cases they will be breaking:

I have a slow and expensive Internet connection used by a few people on a few different devices, I use a proxy-cache to improve page load times and reduce network traffic.

I am a parent, and while I try to be present whenever the kids use the internet, I run a proxy-filter (e.g. DansGuardian) to prevent them from stumbling across less suitable sites.

I am a service provider, and I use a transparent proxy to cache large files downloaded from international sites. This saves me about 10% of my running costs.

I am a service provider provoding internet access with high input costs, in order to provide reasonably-priced services I have quota-based products. In order to be friendly to my customers and avoid them incurring over-use charges, I inject JS notifications at various thresholds. With only HTTPS, I will just have to wait until they are over quota and then block all HTTPS traffic and hope I can redirect some HTTP traffic to a page informing them that they are over quota.

I am a security engineer for my company, for various reasons we need to be able to inspect http traffic (prevent users from visiting malicious sites, enforce productivity controls etc.).

Sure, there are technical means around some of these challenges (e.g. devices that ship with/use CA certs and dynamically generate SSL certs to MITM the traffic), but this initiative is just going to increase costs for everyone.

And who will benefit? Well, most of the main sponsors of Let's encrypt. Cisco will be selling you more network equipment that can MITM SSL, Akamai will get more business as ISPs will not be able to cache on their own and content owners will have to pay Akamai instead.

Maybe some affected parties will start blocking Firefox (or block ssl upgrade checks), or some service providers may start charging Firefox users more.

I am a supporter of open source and have used Firefox as my primary browser since before the 1.0 release, but some of the supposed security braindeadness has made life more difficult, and this is just another example, and may be the one that forces me to change to a web browser, instead of an HTTPS-only browser.

Comment: Re: systemd sux (Score 1) 440

by buchanmilne (#49566669) Attached to: Debian 8 Jessie Released

First, most modern Linux systems come without an inetd or xinetd, because they have no services which aren't supplied by long-running daemons.

Every modern Unix-like system has inetd or xinetd available, many install one of them by default.

The service we require xinetd for on every production server is: Netbackup's bpcd.

Second, inetd won't listen on things it doesn't need to listen on, let alone xinetd.

# readlink -f $(which init) /usr/lib/systemd/systemd
# netstat -plant|grep systemd

How is systemd any different?

Comment: Re: systemd sux (Score 1) 440

by buchanmilne (#49566655) Attached to: Debian 8 Jessie Released

(x)inetd does not control what it attaches, the user does and via plain-text files that are in easy to find standard locations.

# systemctl status rsyncd.socket
rsyncd.socket - Rsync Server Socket
      Loaded: loaded (/usr/lib/systemd/system/rsyncd.socket; disabled)
      Active: inactive (dead)
      Listen: [::]:873 (Stream)
  Accepted: 0; Connected: 0
# cat /usr/lib/systemd/system/rsyncd.socket
Description=Rsync Server Socket



What is this, a non-text file? How is systemd controlling this, any more than xinetd was?

Comment: Re: SystemD added? (Score 1) 493

by buchanmilne (#49549885) Attached to: Ubuntu 15.04 Released, First Version To Feature systemd

No, because you would have already fixed the /etc/init.d/zfs file

If you can fix it in /etc/init.d/zfs, you can just as easily copy /usr/lib/systemd/system/zfs.service to /etc/systemd/system and fix /etc/systemd/system/zfs.service

instead of googling "systemd dependancies editor"

Surely you have vi or emacs or nano or pico or something available, with which to add a Requires entry (see systemd.unit(5)) to the zfs service unit?

Comment: Re:systemd, eh? (Score 1) 493

by buchanmilne (#49549761) Attached to: Ubuntu 15.04 Released, First Version To Feature systemd

Maybe mr smug, you can tell me where on earth the ACPI events from the sleep key are going and why SystemD refuses to pass them on anywhere sensible.

The sleep button works perfectly here on a system running systemd and KDE. Maybe you have a problem somewhere else.

Because I can't debug problems when they arise easily. That makes it pretty inferior to me.

The only seemingly valid complaint I have seen is that systemctl doesn't provide the exist process. But, this is only the case for Type=simple (yes, the default type), where you probably want Type=forking or Type=oneshot.

Comment: Re:systemd rules!!! (Score 1) 493

by buchanmilne (#49549665) Attached to: Ubuntu 15.04 Released, First Version To Feature systemd

I've had to to train all of my junior admins on how to use strace. That took me quite a bit of time, and it takes them a lot of time to go through the huge log files that creates just to find the error string that systemd swallowed. I don't dispute that systemd is better when you have complex dependencies, but it sucks when a unit won't start and it gives you no clue as to why.

You may have been better off reading systemd.service(5), but junior admins should be taught how to use strace regardless ...

I think part of the problem is that sysvinit is basically feature-less, and for a running system actually does nothing (it is initscripts that does this), and so people are used to just having the entire system run by scripts with no useful features (e.g.doing something different with stderr than leaving it to the controlling terminal, letting the current user pollute the environment and thus never have consistent starting of services etc. etc.).

Comment: Re:systemd rules!!! (Score 1) 493

by buchanmilne (#49549657) Attached to: Ubuntu 15.04 Released, First Version To Feature systemd

If you had correctly used Type=oneshot, you wouldn't have been in the dark and would have seen this on the terminal:

# systemctl start broken_systemd
Job for broken_systemd.service failed. See 'systemctl status broken_systemd.service' and 'journalctl -xn' for details.
# systemctl status broken_systemd -l
broken_systemd.service - Broken systemd example
      Loaded: loaded (/etc/systemd/system/broken_systemd.service; disabled)
      Active: failed (Result: exit-code) since Sat 2015-04-25 07:53:07 SAST; 26s ago
    Process: 7880 ExecStart=/root/ (code=exited, status=1/FAILURE)
  Main PID: 7880 (code=exited, status=1/FAILURE)

Apr 25 07:53:07 HOST[7880]: Example systemd service
Apr 25 07:53:07 HOST[7880]: Error that should not be thrown away
Apr 25 07:53:07 HOST systemd[1]: broken_systemd.service: main process exited, code=exited, status=1/FAILURE
Apr 25 07:53:07 HOST systemd[1]: Failed to start Broken systemd example.
Apr 25 07:53:07 HOST systemd[1]: Unit broken_systemd.service entered failed state.

Just because sysvinit couldn't do anything useful with stderr from a one-short service (and leave it to the controlling terminal to do something with it) doesn't mean systemd shouldn't. Logging it, and informing the user that the job didn't start and where to see more information is much more useful.

Comment: Re:Neural Nets (Score 1) 75

by fflewddur (#49443201) Attached to: FTC Creates Office Dedicated To "Algorithmic Transparency"

Absolutely right, though there are still useful things that could be explained: the types of inputs the algorithm accepts, the range of outputs it can potentially give, the model used, etc. In the '90s researchers experimented with building more scrutable models (like decision trees) using a neural net as the training source, with encouraging results [1], but I think the work languished when neural nets went out of fashion.

More importantly, I think this showcases how opaque learning systems (while potentially powerful) may not be appropriate for circumstances when people need to know *why* the system reached a particular conclusion. Predictive accuracy should not be the only metric of concern when developing a machine learning model; comprehensibility of the decision process also needs to be taken into account.

[1] Craven, M. W. and Shavlik, J. W. (1997). Using neural networks for data mining. Future Generation Computer Systems, 13:211–229.

Comment: Re: This whole issue needs to be buried (Score 1) 365

And why is it that the woman is the one to take care of children?

I don't know if you have noticed that men and women are different. Women happen to be more suitable for taking care of babies because they can do one thing men can't: breastfeed.

At least, that is one of the reasons why my wife took 6 months maternity leave, and luckily in those 6 months our situation changed so that we could manage without her working. And we have had more children since, so she is still at home.

Of course, we do realise that she will have fewer years of work experience when/if she returns to work, and this also not earn what she would have if she had continued working. However, the investment in our kids/family life is worth it.

Lavish spending can be disastrous. Don't buy any lavishes for a while.