About this collection
The Signal, the Santa Clarita Valley’s daily paper, on the web from 3 February 1999 to October 2016 (the week of 3 October). 43,371 articles are held.
The first appearance, and the provable start
The paper’s first appearance on the web in this record is a capture of the-signal.com taken on 1 January 1997. It is not an article page. It reads The Signal web site is currently in development.
and links to its on-line classifieds, a guestbook and an e-mail address. The collection does not start there, and this page shows it so a reader can see why.
The 1997 capture
9b3a4afb5b54e68558b14538df8f9df1025ca5a770c06092e8201be824f160c1SHA-256 of the bytes the archive holds, recorded at intake. A SHA-1 of the same bytes, taken at the same time, equals the Wayback Machine’s own digest for this capture: ZMWPNUNYFHOGHSPO5SYW3X6VQVBLOM4J.The provable start is 3 February 1999: the first capture that holds an article. The story it holds carries its own date, 12 June 1998. Between the two, the captures hold a splash page and nothing with an article in it. The first captured front page with headlines is dated 12 October 1999.
- 3 February 1999, the first article capture: SHA-256
b05fb2dc780bfb7370f39c4cee47a2c35747dd3eb19755ed0309ab8aaf0c2946 - 12 October 1999, the first front page with headlines: SHA-256
598875fad7523e8350bcb6e467184c108fb376bc8f7515a24ba726846399807d
Three domains, one paper
The Signal published under three host names in turn. They are the provenance of each record, not divisions of the collection: an article is one record whichever hosts carried it, keyed by the Morris article ID the paper’s system gave it.
- the-signal.com, from the first captures. Its date-file, ViewStory, displaystory and
/news/article/pages exist on this host alone. - signalscv.com, live under the old Morris system from late 2012.
- archive.signalscv.com, first captured on 11 October 2016, 8 days after the old system’s last page on the main domain. It holds the old system’s front page frozen at about 1 October 2016.
Three further subdomains carried the same article systems at another address, and their articles are collected with the rest: oldsite.the-signal.com (the displaystory system), newmedia.the-signal.com (the /news/article/ system) and archive.the-signal.com (the date-file desks). Of the articles held now, 950 from oldsite.the-signal.com, 99 from newmedia.the-signal.com and 70 from archive.the-signal.com.
The old system’s last page on signalscv.com was captured at 3 October 2016, 17:05 UTC; the first page rendered by WordPress was captured on 4 October 2016, 15:52 UTC. The collection ends at that change.
The record lines the subdomain up with the cutover. That is an inference from its timing and its content: no capture says why the subdomain was made. The paper’s sale in June 2018 is not dated by any capture, and nothing in the record ties it to the subdomain.
What is kept
This collection keeps articles The Signal originated and articles about the Santa Clarita Valley: local news, and the paper’s editorials and commentaries about the valley. Articles on national politics are dropped. Where the line falls between the two is the editor’s ruling, made after reading the borderline cases.
7,520 Associated Press news articles that were not about the Santa Clarita Valley were removed on the editor’s ruling of 16 November 2026: they are not unique to this paper, and are probably everywhere on the internet. 68 news articles from other wire services were removed under the same ruling, on the same terms. 90 syndicated columns, each carrying its syndicate’s credit line, were removed on the editor’s ruling of the same day: the collection keeps what is unique to the Signal. Their captures stay in the holding. Local columns, commentary from the wire services, and articles naming the Valley in their dateline, headline or opening were not removed. 8,046 records are awaiting the editor’s ruling. They are counted here and not shown. The machine’s labels on them: 7,743 borderline, 18 clearly-local, 285 clearly-national.
Hosts left out. Four of the paper’s subdomains carry no Signal articles, and are not collected. The counts come from the Wayback Machine’s listings of each host, by address; their pages were not opened, except the community blogs below.
community.the-signal.com: 4,530 distinct addresses, 3,902 of them pages that answered when first captured, none shaped like an article. A reader platform (log-in pages, reader blogs, events, member profiles, galleries).community.signalscv.com: 3,798 distinct addresses, 3,160 of them pages that answered when first captured, none shaped like an article. The same reader platform on the later domain.marketplace.signalscv.com: 2,123 distinct addresses, 2,099 of them pages that answered when first captured, none shaped like an article. A business directory and classified ads.newmedia.signalscv.com: 142 distinct addresses, 24 of them pages that answered when first captured, none shaped like an article. Broken asset paths, a comments-policy page and section stubs.
Community blogs. The reader blogs on the community subdomains are left out; a blog carrying the paper’s staff title is kept as Signal content at an odd address. The staff title is the only test the pages allow, so a staff member who blogged without it would be missed. Of the blog pages held so far, 58 carry the staff title and 996 are reader blogs. Blogs first captured after the October 2016 cutover are held but are not records, staff title or not: 248 blog posts were first captured after it. These pages print no post date, only a Latest Activity
date, so that exclusion rests on the date of first capture, which is not the date of writing. A held page whose Latest Activity date falls on or before 3 October 2016 is kept in the span. Most of the blogs are not yet pulled, so these counts are of what is held at this build.
Republished after the cutover. 293 WordPress posts on signalscv.com, first captured after the October 2016 cutover, carry an address date before it. Compared with the old site’s headlines: 209 are republications of an article the archive holds, and each is listed on that article’s record as another copy, not held. The other 84 match no held article; they fall outside the collection’s span, which ends at the cutover, and are counted here, not shown. The WordPress pages themselves are not held.
Each record carries how it was classified: the label, the classifier and its version, and the signals that fired. That label is a machine’s judgment, recorded as one so it can be revisited.
Publication dates
The collection’s span is about captures: from 3 February 1999, the first capture that holds an article, to the cutover in the week of 3 October 2016. The articles inside that span carry publication dates from 1995, because the earliest captures hold articles published before them: a columnist’s archive captured in 1999 holds columns from years earlier. The chart on the front counts 43,268 articles, by year, from 1995 to 2016.
24 articles carry dates before 1995, each one the date its own page and address give: 23 of them from the paper’s history pages, which reprinted articles with the dates they were first printed. They are listed on their own page, Published before 1995.
79 articles carry a date the archive questions: the page prints one date and its address another. Each is filed by its address, shows both, and is listed on its own page, Date in question, not under a year.
Hosts of origin
Each record’s text comes from one capture on one host, named on its page. By host of origin, the 43,371 records held: the-signal.com 26,510·signalscv.com 14,553·archive.signalscv.com 1,185·archive.the-signal.com 70·newmedia.the-signal.com 99·not recorded 4·oldsite.the-signal.com 950.
One page per article
Where an article survives on more than one host, its page is built from the copy with the most post text; ties go to the copy with the most comments, then the most images, then the latest capture. Words are counted with every e-mail address counted as one token, because the archive host hides addresses that the earlier hosts print, and that difference never decides which copy becomes the record. Each page states which rule decided, and lists every other copy with its hash and what that copy alone holds.
The links that no longer work
This archive’s other collections cite 3,864 distinct Signal addresses. Of those, none still resolves. Measured on 12 November 2026: 3,364 answer with a parked-domain page, 429 answer with a “not found” error, 10 answer with a “forbidden” refusal; 61 have no answer in DNS and are recorded as unresolved, not as gone.
Old addresses on the live site
Old article addresses answer that the page was not found. They do not misroute. On 12 November 2026 the archive asked signalscv.com for 1,336 old article addresses, from seeded random samples and from the articles this archive cites: 1,336 answered not found, and none handed the reader a different page.
Other old addresses do. About 14% of the old site’s 46,356 pages silently redirect to an unrelated post: roughly 6,482, between 6,270 and 7,016 at 95% confidence, measured from a stratified sample of 3,091. They are chiefly its section pages, author pages and preview pages (5,352 of them by that estimate), not its articles. In the sample: section pages, 400 of 400 checked; author pages, 393 of 394 checked; article preview pages, 400 of 400 checked.
Of the 430 signalscv.com addresses this archive’s own collections cite, none misrouted.
Timestamped
The Signal holding’s acquisition manifest as of 13 November 2026 (10,499 fetch records, 10,449 held files; SHA-256 4c8fba95efbda9a9…) was submitted to the OpenTimestamps calendars on 13 November 2026; its proof records an attestation in Bitcoin block 975,654 which the archive has not yet checked against a copy of the Bitcoin chain. Each proof and the manifest it covers are kept in the archive’s records, so anyone can check that the files have not changed since.
