Jump to content

arabellajones

Resident
  • Posts

    561
  • Joined

  • Last visited

Posts posted by arabellajones

  1. OK, that all explains the behaviour of the concurrency numbers for Tuesday and Wednesday. There's a very regular pattern, every day, a consistent shape to the curve. A restart shows a very large dip in concurrency (bots, I suppose) Nothing much on Tuesday, a typical smaller dip on Wednesday.

    I get the data from here: softhyena.com but it only shows the detailed concurrency for the last 24 hours.

  2. As with any other computer hardware, these days, SSDs are not always what they seem. The big fake hardware has been GPU cards, and there are SSD brands which I have never heard of before. I have been careful, and I do get decent results for SL caching. There are a lot of traps in how the caching works, but the bulk of the data arrives at a rate limited by your internet connection.

  3. 9 hours ago, animats said:

    Given the limitations of the Bellissaria rail system, maybe most rail vehicles should be trams, rather than multi-car trains.

    Wouldn't argue with that, but some of those third-party mesh sites explicitly ban use in Second Life. Even without that, there can be LI problems. So be careful.

    • Like 1
  4. I did a quick check, and while there is now an SLRR Rez zone in Randelsham Forest, which is good I didn't, see any other recent changes between there and Bridal Path. In some ways the lines east of Hawkesblood still seem a long way short of the Chalet line.

    • Like 1
  5. 17 hours ago, Mazidox Linden said:

    That's odd, it's working fine for me on both my work machine and personal machine. You might want to try a force refresh in case it's being cached on your end. If that doesn't work, please do let me know.

    CO₂ levels are good, but we may need to adjust the template we use for SLS rolls. I'll check with the folks on the blog-side, thanks!

    You announced the permanent change of the roll-out/restart window back in August.

    Announcement of Permanent Change

    The ability of Linden Lab to run a drunken party in an alcohol production facility is being repeatedly called into doubt.

    • Like 1
  6. I was able to see the effects on log-ins of the roll-out, using the data at https://api.secondlife.com/datafeeds/homepage.txt (the last four lines are what matter). I use a shell script. I may work out one that gives me a graph, but running my desk-top 24/7 to get the data doesn't look worth it. Since I don't use Mac or Windows, I could set up a Raspberry Pi to collect the data, but I have bigger things to worry about.

    The page currently shows three types of data, "signups" (not updated for a year now), "exchange_rate" (updated every 15 minutes),. and "inworld" apparently a live figure). Timestanps are provided in SL time and standard Unix (count of seconds since 00:00:00 UTC on 1 January 1970).

    I don't know if the problem with the old 10-day-uptime trigger for restarts has been fixed, but I frequent a region that has had problems in the past, and there's nothing yet apparent, despite the last restart being in August. Try not to surprise us, some of us are already snug in our beds when the news is released.

  7. 1 hour ago, panterapolnocy said:

    I'd assume it's a web browser cache issue on arabella's side rather than any moron out there. I'd suggest attaching a screenshot of the "new layout" after clearing the cache and reloading the page - or just trying a different browser.

    I think you may be right, but what an earth happened to the page to mess it up?

    Go look at the SL Viewer and you have to wonder at how they cope with the variation in human vision. There are other pieces of software which share the same problem of dark, low-contrast, settings with little or no way of changing it.

  8. I see we have restarts announced for the RC Channels,  at 6am on Wednesday. It's on the Grid Status.

    Meanwhile, the Linden Lab mushroom farm is flourishing. If you got here, you probably saw the horrible new web page. I have words for it, which should not be used in p[olite company.

  9. Huge header section, with a grey background and low-contrast text, I struggle to read what the links presented are, have to scroll down a long way to get to a list of the individual forums, which is hugely extended by blank space and extra info about each forum (maybe useful info). At least when I get into a forum I am seeing the familiar header design, which works and uses readable colours.

    I am trying not to rant about the layout, but what brain-dead moron of a Linden chose such an awful colour scheme?

  10. I can't give any specific advice, but the GeForce 8400 GS is a very old piece of hardware, maybe still usable in a business/office environment, and I'd wonder if the processor can do the things that an SL Viewer needs. I'd check whether the same happens with an alternative up-to-date viewer. The current Firestorm is pretty close to the SL Viewer.

    Updating video cards can get messy; power to the card, the type of video output, and a combination of fake hardware and the chip shortages. Check the benchmarks, relatively ordinary current cards have 20 times the speed while using less power.

  11. 32 minutes ago, arabellajones said:

    Anyway, some pretty pictures to inspire you, and some idea of the options. To get the right proportions on standard SLRR, with the broad gauge, these coaches would be around a nominal 30m long. That suggests as much as 45-degree change in direction in the overall length. I don't think a simple guide sensor at the vehicle sensor would work. Some sort of sensor at each bogie position, and applying the mean of the two positions and directions might work.

    I goofed here. Many of the passing loops  have an S-curve at the switch, so the guide prims at the bogie positions could be parallel. The mean of the positions would work for the position of the centre of the vehicle, but you would need to calculate the vector to get a good direction. The angles might not not be small enough to use the small-angle approximation, but the error may only show briefly. The SLRR gives 8m distance between parallel tracks so I think the approximation should work.

    Points_ani.gif.8a7699617bbdb370b5543965bd6b16f7.gif

    I think I am going to go and have a quiet headache, "llSensor does not detect objects or agents across region boundaries"

  12. I know some of the people reading this are rail geeks, but some aren't. So a few pictures of things that could be handled as articulated vehicles

    1467430819_Wisbech__UpwellSLRR_001.png.202b16f89bd153d6aaf91ad92c6f190b.png

    This is on the Heterocera line, a Y6 Tram Engine with a passenger coach. The history is a bit complicated. but the tram engine is simpler to make a model, because all the moving parts are hidden so as not to scare the horses. The locomotive and coach are similar enough size that the basic SLRR/VRC method of putting the vehicle centre on the Guide prim works.

    SR-Q1-class.jpg.d7175f92bdd2322f2b89c031bfd5e805.jpg

    This is a tender locomotive, and while the proportions are a little different, the tender is not so different from the coach of the first picture. The wheels and coupling rods need animating, but this isn't so complicated because of the inside cylinders. This general type of locomotive was very common across Britain, though this is a WW2 design made to use the minimum of materials. While many earlier designs had more open cabs, the old two-vehicle approach might not work well.

    GWR-Autocoach.jpg.40e15c323affb346008450096b61f4b4.jpg

    This is an Autocoach train, and this would be where things get tricky. It's running right-to-left, not the usual flip we get from the VRC script, and the driver is operation the train from the far end of the coach to the locomotive.  But it's the length of the coach which is awkward. It's a 70ft long bogie coach, and that is so much bigger than the tank locomotive. I am not sure it would be practical.

    GWR-railcar.jpg.7317b48dcfa7eebd3419f418f67c1523.jpg

    This would be a better answer for passenger trains on Bellisseria, the basic idea was being tried in several countries before WW2. This is a diesel streamlined railcar of the GWR, and while the guide-prim details for such a long vehicle could be awkward, it's still a single vehicle. At least initially, the bogies were shrouded. These were fast, rather opulent, trains. Some of the early ones had a buffet/bar. Unlike the autocoach solution it wouldn't much matter if you used the old-style Flip to reverse travel.

     

    Anyway, some pretty pictures to inspire you, and some idea of the options. To get the right proportions on standard SLRR, with the broad gauge, these coaches would be around a nominal 30m long. That suggests as much as 45-degree change in direction in the overall length. I don't think a simple guide sensor at the vehicle sensor would work. Some sort of sensor at each bogie position, and applying the mean of the two positions and directions might work.

     

  13. 4 hours ago, Qie Niangao said:

    Yeah, this land setting (disallowing Object Entry) means that only sat-upon objects can traverse the path, which for my interests renders railway throughout Bellisseria ornamental at best. Among the problems are multi-unit trains, as you mention, which can't really be "fixed" in any practical way, inasmuch as all the units would need to be linked in a single assembly, requiring a whole lot of geometry calculations, especially complicated for curves at or near region borders. That's assuming that such a complex assembly of links can make it across the border at all.

    Anyone who's looked at the VRC train scripts knows they could be much simplified, in part because they antedate many "efficient script" functions for manipulating link properties. I've sometimes been tempted to tidy-up, but I always worry they may unwittingly depend on accidents of timing.

    That complexity is why I suggested a tender locomotive. A GWR Autocoach could soon get too complicated.

  14. The problem with Hawkesblood seems to be the naming of the guide prims at either end of the rez zone. It may be related to the signals on either side of the station. They have scripts. I have found I have a copy of the v2 signal script system, and the set-up looks complicated.

    • Like 1
  15. I have now being able to check the increased SENSOR_SEEK_ARC_DEFAULT at the boundary between SSPE1124 and SSPE1125 which has been one of the black spots for me. The northbound crossing actually worked for me with a value of 30.00 and I have not yet seen other problems. But I am not sure because I have seen that Squeaky Mole has added two extra guide prims, the long ones you see in the picture. It looks as if they would reduce any sensor angle problem.Problem-Sim-Crossing_001.jpg.f752a5ef49465344392db26879c245ee.jpg 

    The rate of change for the short prims is, as near as I can tell, marginal for a 25-degree sensor arc. The SLRR specification for curves is not all that clear, but I read it as saying more than 15 degrees in 10m needs caution. Trouble is that the region crossing process adds delays to scripts so the critical sensor ping could completely miss.

    • Like 1
  16. I have made several test runs with a locomotive of mine on the SLRR-compatible rail lines of the Bellisseria continent. Much is still unfinished, but usable. The main problems are the limited number of rez-zones and a few unreliable region crossings. There's also a difference in the Land Permissions which means multi-unit trains need to be built as a single vehicle. The follower-vehicle method used by many scripts fails. Anyone building a tender locomotive needs scripts which fix this

    The main section is a loop running through Pugwash in the west, with a very long spur to Bridle Path in the north-east. There are two branches running to the south coast, both with nearby docks, and a junction at Gully Wash for a line to the north coast which is clearly expected to be extended, the same sort of sharp cut-off of scenery you get if the adjacent region is down.

    There is also a Chalet sub-continent line, mostly following the Big River, from Hydrant to Lavon, again with likely extensions at each end.

    I have found that Land Impact can be critical in getting reliable operation. Many of the older VRC and Hobo freebies are very old, prim-heavy tech. I have managed to get a mesh version of a prim-based model from LI=27 to LI=8, and it has improved running.

    Some places have rather suspect track design, features such as sharp curves at region crossings (there's a bad one just south of Gully Wast). I have a suspicion that it needs a change to the locomotive script to fix. I am going to increase the SENSOR_SEEK_ARC_DEFAULT but the paucity of rez zones is going to make testing tedious. Since there are similar failures on dead-straight sections of track, in mid-region I shall also try increasing SENSOR_SEEK_RANGE_DEFAULT

    image_2021-08-22_073600.png.db50d1b2e268648beb57917ffbc05c9d.png

    I am working, fitfully, on a model of this, to allow high-speed passenger services on the Bellisseria lines.

    34354793560_129ef889f7_b.jpg.ae9d9c3425ba287574667c2632799ec6.jpg

    I will note that many of these Bellisseria lines run through regions with an SSPE label, which Linden Lab regard as officially unfinished. But the Chalet section has blatantly unfinished track (look at Lavon with visible guide prims) in named regions.

    • Like 1
×
×
  • Create New...