A public sky slideshow is a sequence of stills. The scientific test is whether those stills play in capture-time order. Autoplay that shuffles, repeats a random hero, or starts in the middle is a broken observation sequence, even if the fade looks polished. Visitors who followed /nslide/html5slide1.php were opening a public HTML5 slider on the historical station host. This page keeps that address. It does not run a slider, does not rehost stills, and does not ship a JavaScript plugin.
A related QA problem lives on a different host and a different path: cumulus.tnetweather.com/test/nivotest is a test harness for Nivo-style order. This URL is the public slideshow. Production visitors should never be pointed at /test/. The order rule is the same. The audience is not.
Historical context
Station templates in the HTML5 era replaced Flash and some jQuery galleries with a div of <img> tags and a small script that advanced them. Names such as nslide and html5slide1 mark that family: a numbered public slide page, often fed by webcam stills or a day’s FTP drops. Incoming links treated the URL as the slideshow. The original player and its folder of JPEGs are not rehosted.
The station webcam is the instrument: interval, field of view, timestamp. The gallery index is the archive catalog. The slider sits downstream of both. Given a list of stills that already have times, does the public page present them as a chronological sequence?
A slider is a derived display
Each still is a historical image at a stated time, if the time traveled with the file. The slider is a derived display: it orders those images and spends some number of milliseconds on each. If the order is wrong, cloud motion appears to reverse. A shower “happens” before the clear noon that was actually earlier. Visitors remember the animation, not the filenames.
Playback duration is not observation interval. A three-second pause between fifteen-minute stills is a viewing choice. The caption must still show the fifteen-minute gap. Overlay clocks, if present, must match the still’s capture time, not the browser’s clock, and not the slider’s setInterval.
HTML5 made this easier to do well and easier to do badly. A native <img> list with Next and Previous, sorted by capture time, is an honest sequence. A carousel that clones nodes, preloads in file-size order, or sets shuffle: true for “interest” is a false time-lapse. A <video> built from stills can be honest if the encoder received frames in time order and the poster frame is the first observation, not a random sunset.
What to require of a public sequence
The public page is not a unit test, but it still has pass/fail rules.
Source order. The HTML, JSON, or playlist lists frames in nondecreasing capture time. Filename sort is acceptable only if the filename encodes a sortable timestamp and a camera ID. slide12.jpg then slide2.jpg is lexical order, not time order. Gallery “last uploads” order is a website event, not an observation series.
Player order. After load, the visible sequence matches that list. If the library copies nodes into a wrapper, the wrapper must not sort by upload mtime, file size, or randomness.
Start. The first frame is the earliest in the published window, or a documented start that the caption states. A random start is a failed weather sequence.
Loop policy. Looping to the earliest frame is acceptable if the caption says the sequence repeats. Looping into a second camera’s folder is not. Mixing north-sky stills with a yard camera without a title card is a silent FOV change.
Stale and missing. If a slot in the interval is missing, skip it or show a gap card. Do not stretch the previous still and let the overlay clock lie. Do not pull current.jpg into a historical sequence; that alias duplicates the newest frame and breaks uniqueness.
If capture times are missing, the slider cannot be an observation product. It can be a gallery. Galleries should not be titled as time-lapse. Cite individual files with capture time and identity instead of animating them.
Autoplay is allowed only as ordered walking
Autoplay is the usual way a home page “looks alive.” A shuffled sequence with a fade still looks alive. On a public slideshow, autoplay is permitted when it walks the time-ordered list and the caption stays in lockstep. Autoplay coupled to a random effect that also randomizes content is not permitted.
Pause on hover is a viewing convenience. It must not freeze the displayed clock at “now.” The still’s time stays the still’s time.
Do not trust FTP mtime to build the playlist. A batch upload stamps every file with the same second. Do not trust the slider’s start index after someone prepends latest.jpg as a copy of the newest still.
The test harness on /test/nivotest is where an operator asserts order with a fixture set. The public /nslide/ page is where visitors see the result. Do not expose the fixture directory, a glob of /webcams/, or a query parameter that loads arbitrary paths. That is file disclosure, not a slideshow.
What this page is not
It is not a plugin recommendation. Leftover sliders from the jQuery and early-HTML5 era are often unmaintained. XSS and dependency issues are a reason to keep old copies off a public URL, not a reason to advertise a brand. Newer sites can use CSS, native video, or stacked images with buttons. The order rule does not care.
It is not a live webcam. This URL does not proxy a camera and does not restore Mesa imagery.
It is not a cloud classifier. Animating JPEGs does not produce a METAR. The only meteorological claim a sequence may make is: these frames are in the order of their capture times, as labeled.
It is not the Cumulus Nivo test. Different host, different path, different audience.
Practical checklist
- Playlist is sorted by capture time; lexical filename sort only when names encode time and camera ID.
- Autoplay walks that list; random start and content shuffle are off.
- Captions show time and timezone on the current frame; pause time is not the observing interval.
- One camera and one FOV per sequence, or an explicit title card at a change.
- Gaps are gaps;
current.jpgaliases stay out of the historical list. - Production does not point at
/test/harnesses or directory globs. - If times are unknown, do not call it a time-lapse.
Modern relevance
CSS carousels and design-system “hero sliders” recreated the old failure with new class names. A weather sequence is not a marketing banner. The banner may shuffle. The sky may not.
Operators who still publish interval stills from the station webcam can offer a day’s sequence as HTML5 without pretending it is video. The archive remains the observation log. The slider is only a player.
TNET does not run html5slide1 at this path. Anyone who followed /nslide/html5slide1.php for a slideshow gets the sequence contract that slideshow required.
TNET research bridge
Time-ordered imagery is an evidence-handling problem: identity of the camera, time of each frame, and no silent shuffle. A still is observed. The animation is derived. That vocabulary is how the service works. Source identity and quality control for the files in the playlist are data sources, quality controls, and methodology. The weather-station hub lists webcam and gallery neighbors.
Sources
- CIPA DC-008 (Exif) — capture time versus file-change time (last checked 13 August 2026)
- MDN: The Image Embed element