[Genma Saotome]

Once again the (unstable) update failed and trashed my installation. I have to go to the download page and get the whole package again and that experience leads to this post: It took 18 hours to get the install package. I have a download speed alleged to be 46mbs, usually mid 30's, so the problem is not on my end.

18 hours is the norm in my experience (having had the update trash my install several times).

Something really isn't right, esp. when contrasted w/ free, reasonably fast services from Dropbox and/or Google Drive.

[James Ross]

Genma Saotome, on 24 July 2017 - 10:27 AM, said:

Once again the (unstable) update failed and trashed my installation. I have to go to the download page and get the whole package again and that experience leads to this post: It took 18 hours to get the install package. I have a download speed alleged to be 46mbs, usually mid 30's, so the problem is not on my end.

18 hours is the norm in my experience (having had the update trash my install several times).

Something really isn't right, esp. when contrasted w/ free, reasonably fast services from Dropbox and/or Google Drive.

Which version are you downloading and from where?

• Downloading the stable version from http://openrails.org/download/program/ takes 15 seconds here.
  
• Downloading the testing version from http://openrails.org/download/program/ takes 10 seconds here.
  
• Downloading the unstable version from https://james-ross.co.uk/projects/or takes about 8 seconds here.

The above times were all from Edge, but I also tried Chrome, Firefox, and Internet Explorer and they gave very similar times.

If you are downloading from openrails.org, please try the alternative link in the confirmation page and let me know if that performs differently.

The statistics for the speed of downloads from openrails.org shows that 46% of downloads complete within 5 seconds, with just under 20% taking two minutes or more (downloads often have a long tail). Note that the right-most bar is the tail, i.e. everything that took longer than 120s is included within it.

https://james-ross.co.uk/temp/orts_149.png

I also checked the number of downloads via the direct link vs PHP (the latter being the default), and there is a reasonable minority (just under 23%) of such downloads. I am curious at the size of this, so we might try some changes to the site over the next few weeks to see what happens.

https://james-ross.co.uk/temp/orts_150.png

If you, or anyone else getting slow downloads from openrails.org, has any download accelerators or download-related extensions, please try with them disabled so that any compatibility issues can be identified.

[thegrindre]

I have the slowest internet available and my download speed is in access of 200kb/s or so. Takes about 5 minutes to DL and I have to use the alternative link.

http://www.elvastower.com/forums/public/style_emoticons/default/oldstry.gif

[Genma Saotome]

Using copy link I see this: http://openrails.org...program/#modal1


IIRC the download in the basenote was kicked off around 10PM my time last night (I'm GMT-8) and completed about 12:30PM this afternoon. So only 13.5 hours (I had guessed longer). I've just downloaded it again... it's coming in at an average speed of 378 KB/s and took a minute 50 seconds. In my experience that's lightening fast for this download.

Doing another download right now it's paused at 78%. AFAIK I did not do anything to cause a pause of the download. Oh... now it says failed.

One more time.... 333 KB/s over 2 minutes 5 seconds.

When I request an update of a stable version speed seems to be fine but as I've noted it occasionally screws the pooch and leaves the OR directory non-functional.

[Hobo]

I have just downloaded 3 different latest versions at about 395 kb - - 26 - 28 seconds each using your link and no problems so far .

[R H Steele]

I usually download from http://james-ross.co...jects/or/builds, always dependable.
Frequently have problems with primary testing download on OR----.org site. I've kept a log and curiously enough the download always fails at 90%. The speed does not vary much, approx 350Kbps, slower than most other sites I visit - and definitely slower than my provider advertises, but that's to be expected -- normal provider hype.
The alternative site always seems to work just fine when I do decide to use it.
Never have used the auto update, was too undependable, do not mind doing the manual downloads.

[James Ross]

Genma Saotome, on 24 July 2017 - 03:45 PM, said:

Using copy link I see this: http://openrails.org...program/#modal1

Okay, so that's the stable version installer. If you are going to switch the update channel to something else (testing or unstable), you can save a bit of the download by downloading those versions directly [1].

Genma Saotome, on 24 July 2017 - 03:45 PM, said:

IIRC the download in the basenote was kicked off around 10PM my time last night (I'm GMT-8) and completed about 12:30PM this afternoon. So only 13.5 hours (I had guessed longer). I've just downloaded it again... it's coming in at an average speed of 378 KB/s and took a minute 50 seconds. In my experience that's lightening fast for this download.

Doing another download right now it's paused at 78%. AFAIK I did not do anything to cause a pause of the download. Oh... now it says failed.

One more time.... 333 KB/s over 2 minutes 5 seconds.

This is quite strange. Where in the world are you (and anyone else having slow speeds)? I wonder if there's some issue with the hosting and some particular countries more generally - I'll have a look at global performance, if I can.

I've had a look at the Google Analytics for the site, which doesn't measure the download times for files but does record page load times globally, and it doesn't look like any region suffers more than I was expecting; for example, Northern America gets a mean page load time of 3.33s vs the site global mean of 6.38s, while Asia and Australasia suffer somewhat but even the worst sub-continent of Western Asia is only 11.57s.

I can't find any site that'll download a file from a website from multiple worldwide locations and show the speeds; if anyone knows of one, I'm interested, although given the Google Analytics data it might not show up anything.

I'll see if I can hook up GeoIP to the website log files.

Can you compare the normal download behaviour and the alternative link? I'd see how fast the normal download is going after 10 seconds or so, then cancel it, then start the alternative download and see how fast it goes.

Genma Saotome, on 24 July 2017 - 03:45 PM, said:

When I request an update of a stable version speed seems to be fine but as I've noted it occasionally screws the pooch and leaves the OR directory non-functional.

That's very interesting; the updates are not particularly different to the normal download links - stable is a bit smaller as it downloads a zip instead of the exe, while testing and unstable are identical.

[1] Download sizes:

• Stable: 40.5 MB from website
  
• Stable: 34.3 MB from app
  
• Testing: 35.1 MB from website and app
  
• Unstable: 8.99 MB from website and app

[James Ross]

James Ross, on 25 July 2017 - 11:47 AM, said:

I'll see if I can hook up GeoIP to the website log files.

I've got it hooked up and we get...

• Turkey: 6s
  
• Monaco: 17s
  
• ...
  
• United Kingdom: 56s
  
• ...
  
• United States: 69s
  
• ...
  
• India (our top download location): 69s
  
• ...
  
• Hong Kong: 390s
  
• Austria: 1019s

[Mike B]

I have to agree that downloading anything from the ORTS web site is a coinflip. Actually, worse that that - a coinflip on average wins 1/2 of the time. ORTS website download fails over 75% of the time, usually following the pattern described: downloads slowly (though, for me, usually taking about 1/2-3/4 hour for the full installer) then fails at around 90%.

As for updates from within the app, I had similar problems until about 3 months ago, since which time update downloads have been almost always quick (a minute or 3) and reliable. Before that, they would usually fail, and would take the OR directory with them requiring a complete new install.

Downloads of testing/unstable versions from your web site are and have been uniformly quick and reliable.

Yes, I have slow (by world standards) internet service - officially 6, but varies from 1 to as much as 8 mbit/sec. West Coast of the US. Quicker during the day when everybody else in the neighborhood is at work. EDIT: realized that using Downthemall (download accelerator extension in Firefox) does in fact kill all downloads - something at the OR web site is incompatible with it. I've done straight downloads (through the browser itself) that have succeeded, but still with a low probability of success.

Bottom line: once I *do* get a download to complete from the ORTS web site, I immediately archive the file so I won't have to do it again anytime soon.

[James Ross]

Mike B, on 25 July 2017 - 12:45 PM, said:

As for updates from within the app, I had similar problems until about 3 months ago, since which time update downloads have been almost always quick (a minute or 3) and reliable. Before that, they would usually fail, and would take the OR directory with them requiring a complete new install.

If the updater fails during the download stage, nothing should be done to the already-installed files, so this is strange.

Mike B, on 25 July 2017 - 12:45 PM, said:

Yes, I have slow (by world standards) internet service - officially 6, but varies from 1 to as much as 8 mbit/sec. West Coast of the US. Quicker during the day when everybody else in the neighborhood is at work. EDIT: realized that using Downthemall (download accelerator extension in Firefox) does in fact kill all downloads - something at the OR web site is incompatible with it. I've done straight downloads (through the browser itself) that have succeeded, but still with a low probability of success.

Thanks for identifying Downthemall - download accelerators have been a cause of issues for people throughout the decades IME, and I would recommend not using them unless you have a very specific problem (like highly unreliable connection) that you know they help with. The alternative link should work fine with anything, though, as it is a direct download link - so problems with that are something else (i.e. things totally outside our control).